Large vision-language models (VLMs) for autonomous driving (AD) are evolving beyond perception and cognition tasks toward motion planning. However, we identify two critical challenges in this direction: (1) VLMs tend to learn shortcuts by relying heavily on history input information, achieving seemingly strong planning results without genuinely understanding the visual inputs; and (2) the chain-of-thought (COT) reasoning processes are always misaligned with the motion planning outcomes, and how to effectively leverage the complex reasoning capability to enhance planning remains largely underexplored. In this paper, we start from a small-scale domain-specific VLM and propose Drive-R1, designed to bridge the scenario reasoning and motion planning for AD. Drive-R1 first undergoes the supervised finetuning on an elaborate dataset containing both long and short COT data. Drive-R1 is encouraged to reason step-by-step from visual input to final planning decisions. Subsequently, Drive-R1 is trained within a reinforcement learning framework that incentivizes the discovery of reasoning paths that are more informative for planning, guided by rewards based on predicted trajectories and meta actions. Experimental evaluations on the nuScenes and DriveLM-nuScenes benchmarks demonstrate that Drive-R1 achieves superior performance compared to existing state-of-the-art VLMs. We believe that Drive-R1 presents a promising direction for bridging reasoning and planning in AD, offering methodological insights for future research and applications.

One-shot pruning efficiently compresses Large Language Models but produces coarse sparse weights, causing significant performance degradation. Traditional fine-tuning approaches to refine these weights are prohibitively expensive for large models. This highlights the need for a training-free weight refinement method that works seamlessly with one-shot pruning and can efficiently recover the lost performance. To tackle this problem, we propose Efficient Iterative Weight Refinement (EIWR), a lightweight, plug-and-play, and training-free method that refines pruned weights through layer-wise iterative optimization. EIWR achieves efficient weight refinement via three key components: a Global Soft Constraint that eliminates costly row-wise Hessian inversions and expands the solution space; a Historical Momentum Strategy that leverages one-shot pruning priors to accelerate convergence and enhance final performance; and Neumann Series Extrapolation that significantly speeds up per-iteration computation. As a result, EIWR enables effective weight refinement with minimal time and memory overhead. Extensive experiments on LLaMA2/3 and Qwen under different pruning strategies and sparsity levels demonstrate that our method can efficiently refine sparse weights and mitigate performance degradation. For example, on LLaMA2-7B under 70 percent sparsity, EIWR reduces perplexity by 15 percent compared with SparseGPT on the WikiText2 benchmark, with only 1.81 additional minutes of computation and 1GB of additional memory.

Xiaoyu Liu, Yun Zhang, Wei Li, Simiao Li, Xudong Huang, Hanting Chen, Yehui Tang, Jie Hu, Zhiwei Xiong, Yunhe Wang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Reconstructing neurons from large electron microscopy (EM) datasets for connectomic analysis presents a significant challenge, particularly in segmenting neurons of complex morphologies. Previous deep learning-based neuron segmentation methods often rely on pixel-level image context and produce extensive oversegmented fragments. Detecting these split errors and merging the split neuron segments are non-trivial for various neurons in a large-scale EM data volume. In this work, we exploit multimodal features in the full workflow of automatic neuron proofreading. We propose a novel connection point detection network that utilizes both global 3D morphological features and high-resolution local image context to extract candidate segment pairs from massive adjacent segments. To effectively fuse the 3D morphological feature and the dense image features from very different scales, we design a proposal-based image feature sampling to improve the efficiency of multimodal cross-attentions. Integrating the connection point detection network with our connectivity prediction network which also utilizes multimodal features, we make a fully automatic neuron segment merging pipeline, closely imitating human proofreading. Comprehensive experimental results verify the effectiveness of the proposed modules and demonstrate the robustness of the entire pipeline in large-scale neuron reconstruction. The code and data are available at https://github.com/Levishery/Neuron-Segment-Connection-Prediction.

Structured light (SL) is a popular approach for 3D reconstruction. Most SL techniques rely on frame-based cameras and are often not robust in high-speed dynamic scenes. Recently, event cameras have sparked growing interest in high-speed SL imaging, due to their high temporal resolution. The event-based SL enjoys the high-speed data acquisition, however, most existing methods tend to pursue the reconstruction accuracy but sacrificing the computational efficiency, limiting the applicability in real-world scenarios. To this end, we propose E2SL, an Efficient deep network tailored for monocular Event-based SL. Specifically, E2SL comprises three key components: binary embedding lookup table (BE-LUT), spatial context enhancement (SCE), and geometric-prior regression (GPR). Given the input event frame, BE-LUT, which is precomputed and stored, first retrieves the features efficiently. Then, SCE extends the receptive field of the features and captures the spatial context. Finally, GPR conducts the geometric-prior-based tree classification for fast and robust depth estimation. To support training and evaluation, we contribute an event-based SL simulator, which generates a large-scale and diverse synthetic dataset. Besides, we develop an event-based SL prototype and collect a dataset with accurate ground truth for real-world evaluation. Extensive experiments demonstrate that our method achieves state-of-the-art accuracy while maintaining a per-frame reconstruction time of 7.7 ms, meeting the demands of high-speed depth sensing. The code and dataset are available on the project page https://dongxin000.github.io/E2SL/.

In this paper, we tackle the task of blurry video super-resolution (BVSR), aiming to generate high-resolution (HR) videos from low-resolution (LR) and blurry inputs. Current BVSR methods often fail to restore sharp details at high resolutions, resulting in noticeable artifacts and jitter due to insufficient motion information for deconvolution and the lack of high-frequency details in LR frames. To address these challenges, we introduce event signals into BVSR and propose a novel event-enhanced network, Ev-DeblurVSR. To effectively fuse information from frames and events for feature deblurring, we introduce a reciprocal feature deblurring module that leverages motion information from intra-frame events to deblur frame features while reciprocally using global scene context from the frames to enhance event features. Furthermore, to enhance temporal consistency, we propose a hybrid deformable alignment module that fully exploits the complementary motion information from inter-frame events and optical flow to improve motion estimation in the deformable alignment process. Extensive evaluations demonstrate that Ev-DeblurVSR establishes a new state-of-the-art performance on both synthetic and real-world datasets. Notably, on real data, our method is 2.59 dB more accurate and 7.28× faster than the recent best BVSR baseline FMA-Net.

In this paper, we present S2D-LFE, an innovative approach for sparse-to-dense light field event generation. For the first time to our knowledge, S2D-LFE enables controllable novel view synthesis only from sparse-view light field event (LFE) data, and addresses three critical challenges for the LFE generation task: simplicity, controllability, and consistency. The simplicity aspect eliminates the dependency on frame-based modality, which often suffers from motion blur and low frame-rate limitations. The controllability aspect enables precise view synthesis under sparse LFE conditions with view-related constraints. The consistency aspect ensures both cross-view and temporal coherence in the generated results. To realize S2D-LFE, we develop a novel diffusion-based generation network with two key components. First, we design an LFE-customized variational auto-encoder that effectively compresses and reconstructs LFE by integrating cross-view information. Second, we design an LFE-aware injection adaptor to extract comprehensive geometric and texture priors. Furthermore, we construct a large-scale synthetic LFE dataset containing 162 one-minute sequences using simulator, and capture a real-world testset using our custom-built sparse LFE acquisition system, covering diverse indoor and outdoor scenes. Extensive experiments demonstrate that S2D-LFE successfully generates up to 9 × 9 dense LFE from 2 × 2 sparse inputs and outperforms existing methods on both synthetic and real-world data. The datasets and code are available at https://github.com/Yutong2022/S2D-LFE.

As a widely adopted technique in data transmission, video compression effectively reduces the size of files, making it possible for real-time cloud computing. However, it comes at the cost of visual quality, posing challenges to the robustness of downstream vision models. In this work, we present a versatile codec-aware enhancement framework that reuses codec information to adaptively enhance videos under different compression settings, assisting various downstream vision tasks without introducing computation bottleneck. Specifically, the proposed codec-aware framework consists of a compression-aware adaptation (CAA) network that employs a hierarchical adaptation mechanism to estimate parameters of the frame-wise enhancement network, namely the bitstream-aware enhancement (BAE) network. The BAE network further leverages temporal and spatial priors embedded in the bitstream to effectively improve the quality of compressed input frames. Extensive experimental results demonstrate the superior quality enhancement performance of our framework over existing enhancement methods, as well as its versatility in assisting multiple downstream tasks on compressed videos as a plug-and-play module. Code and models are available at https://huimin-zeng.github.io/PnP-VCVE/.

Existing benchmarks for Vision-Language Model (VLM) on autonomous driving (AD) primarily assess interpretability through open-form visual question answering (QA) within coarse-grained tasks, which remain insufficient to assess capabilities in complex driving scenarios. To this end, we introduce $\textbf{VLADBench}$, a challenging and fine-grained dataset featuring close-form QAs that progress from static foundational knowledge and elements to advanced reasoning for dynamic on-road situations. The elaborate $\textbf{VLADBench}$ spans 5 key domains: Traffic Knowledge Understanding, General Element Recognition, Traffic Graph Generation, Target Attribute Comprehension, and Ego Decision-Making and Planning. These domains are further broken down into 11 secondary aspects and 29 tertiary tasks for a granular evaluation. A thorough assessment of general and domain-specific (DS) VLMs on this benchmark reveals both their strengths and critical limitations in AD contexts. To further exploit the cognitive and reasoning interactions among the 5 domains for AD understanding, we start from a small-scale VLM and train the DS models on individual domain datasets (collected from 1.4M DS QAs across public sources). The experimental results demonstrate that the proposed benchmark provides a crucial step toward a more comprehensive assessment of VLMs in AD, paving the way for the development of more cognitively sophisticated and reasoning-capable AD systems.

Non-line-of-sight (NLOS) imaging, recovering the hidden volume from indirect reflections, has attracted increasing attention due to its potential applications. Despite promising results, existing NLOS reconstruction approaches are constrained by the reliance on empirical physical priors, e.g., single fixed path compensation. Moreover, these approaches still possess limited generalization ability, particularly when dealing with scenes at a low signal-to-noise ratio (SNR). To overcome the above problems, we introduce a novel learning-based solution, comprising two key designs: Learnable Path Compensation (LPC) and Adaptive Phasor Field (APF). The LPC applies tailored path compensation coefficients to adapt to different objects in the scene, effectively reducing light wave attenuation, especially in distant regions. Meanwhile, the APF learns the precise Gaussian window of the illumination function for the phasor field, dynamically selecting the relevant spectrum band of the transient measurement. Experimental validations demonstrate that our proposed approach, only trained on synthetic data, exhibits the capability to seamlessly generalize across various real-world datasets captured by different imaging systems and characterized by low SNRs.

We present TAR3D, a novel framework that consists of a 3D-aware Vector Quantized-Variational AutoEncoder (VQ-VAE) and a Generative Pre-trained Transformer (GPT) to generate high-quality 3D assets. The core insight of this work is to migrate the multimodal unification and promising learning capabilities of the next-token prediction paradigm to conditional 3D object generation. To achieve this, the 3D VQ-VAE first encodes a wide range of 3D shapes into a compact triplane latent space and utilizes a set of discrete representations from a trainable codebook to reconstruct fine-grained geometries under the supervision of query point occupancy. Then, the 3D GPT, equipped with a custom triplane position embedding called TriPE, predicts the codebook index sequence with prefilling prompt tokens in an autoregressive manner so that the composition of 3D geometries can be modeled part by part. Extensive experiments on ShapeNet and Objaverse demonstrate that TAR3D can achieve superior generation quality over existing methods in text-to-3D and image-to-3D tasks

Post-training quantization (PTQ) has played a pivotal role in compressing large language models (LLMs) at ultra-low costs. Although current PTQ methods have achieved promising results by addressing outliers and employing layer- or block-wise loss optimization techniques, they still suffer from significant performance degradation at ultra-low bits precision. To dissect this issue, we conducted an in-depth analysis of quantization errors specific to LLMs and surprisingly discovered that, unlike traditional sources of quantization errors, the growing number of model parameters, combined with the reduction in quantization bits, intensifies inter-layer and intra-layer dependencies, which severely impact quantization accuracy. This finding highlights a critical challenge in quantizing LLMs. To address this, we propose CBQ, a cross-block reconstruction-based PTQ method for LLMs. CBQ leverages a cross-block dependency to establish long-range dependencies across multiple blocks and integrates an adaptive LoRA-Rounding technique to manage intra-layer dependencies. To further enhance performance, CBQ incorporates a coarse-to-fine pre-processing mechanism for processing weights and activations. Extensive experiments show that CBQ achieves superior low-bit quantization (W4A4, W4A8, W2A16) and outperforms existing state-of-the-art methods across various LLMs and datasets. Notably, CBQ only takes 4.3 hours to quantize a weight-only quantization of a 4-bit LLAMA1-65B model, achieving a commendable trade off between performance and efficiency.

For a more compatible and consistent high dynamic range (HDR) viewing experience, a new image format with a double-layer structure has been developed recently, which incorporates an auxiliary Gain Map (GM) within a standard dynamic range (SDR) image for adaptive HDR display. This new format motivates us to introduce a new task termed Gain Map-based Inverse Tone Mapping (GM-ITM), which focuses on learning the corresponding GM of an SDR image instead of directly estimating its HDR counterpart, thereby enabling a more effective up-conversion by leveraging the advantages of GM. The main challenge in this task, however, is to accurately estimate regional intensity variation with the fluctuating peak value. To this end, we propose a dual-branch network named GMNet, consisting of a Local Contrast Restoration (LCR) branch and a Global Luminance Estimation (GLE) branch to capture pixel-wise and image-wise information for GM estimation. Moreover, to facilitate the future research of the GM-ITM task, we build both synthetic and real-world datasets for comprehensive evaluations: synthetic SDR-GM pairs are generated from existing HDR resources, and real-world SDR-GM pairs are captured by mobile devices. Extensive experiments on these datasets demonstrate the superiority of our proposed GMNet over existing HDR-related methods both quantitatively and qualitatively. The codes and datasets are available at https://github.com/qtlark/GMNet.

Achieving high-fidelity audio compression while preserving perceptual quality across diverse audio types remains a significant challenge in Neural Audio Coding (NAC). This paper introduces MUFFIN, a fully convolutional NAC framework that leverages psychoacoustically guided multi-band frequency reconstruction. Central to MUFFIN is the Multi-Band Spectral Residual Vector Quantization (MBS-RVQ) mechanism, which quantizes latent speech across different frequency bands. This approach optimizes bitrate allocation and enhances fidelity based on psychoacoustic studies, achieving efficient compression with unique perceptual features that separate content from speaker attributes through distinct codebooks. MUFFIN integrates a transformer-inspired convolutional architecture with proposed modified snake activation functions to capture fine frequency details with greater precision. Extensive evaluations on diverse datasets (LibriTTS, IEMOCAP, GTZAN, BBC) demonstrate MUFFIN’s ability to consistently surpass existing performance in audio reconstruction across various domains. Notably, a high-compression variant achieves an impressive SOTA 12.5 kHz rate while preserving reconstruction quality. Furthermore, MUFFIN excels in downstream generative tasks, demonstrating its potential as a robust token representation for integration with large language models. These results establish MUFFIN as a groundbreaking advancement in NAC and as the first neural psychoacoustic coding system. Speech demos and codes are available at https://demos46.github.io/muffin/ and https://github.com/dianwen-ng/MUFFIN.

Recent works have correlated Masked Image Modeling (MIM) with consistency regularization in Unsupervised Domain Adaptation (UDA). However, they merely treat masking as a special form of deformation on the input images and neglect the theoretical analysis, which leads to a superficial understanding of masked reconstruction and insufficient exploitation of its potential in enhancing feature extraction and representation learning. In this paper, we reframe masked reconstruction as a sparse signal reconstruction problem and theoretically prove that the dual form of complementary masks possesses superior capabilities in extracting domain-agnostic image features. Based on this compelling insight, we propose MaskTwins, a simple yet effective UDA framework that integrates masked reconstruction directly into the main training pipeline. MaskTwins uncovers intrinsic structural patterns that persist across disparate domains by enforcing consistency between predictions of images masked in complementary ways, enabling domain generalization in an end-to-end manner. Extensive experiments verify the superiority of MaskTwins over baseline methods in natural and biological image segmentation. These results demonstrate the significant advantages of MaskTwins in extracting domain-invariant features without the need for separate pre-training, offering a new paradigm for domain-adaptive segmentation. The source code is available at https://github.com/jwwang0421/masktwins.

Controllable 3D indoor scene synthesis stands at the forefront of technological progress, offering various applications like gaming, film, and augmented/virtual reality. The capability to stylize and decouple objects within these scenarios is a crucial factor, providing an advanced level of control throughout the editing process. This control extends not just to manipulating geometric attributes like translation and scaling but also includes managing appearances, such as stylization. Current methods for scene stylization are limited to applying styles to the entire scene, without the ability to separate and customize individual objects. Addressing the intricacies of this challenge, we introduce a unique pipeline designed for synthesis 3D indoor scenes. Our approach involves strategically placing objects within the scene, utilizing information from professionally designed bounding boxes. Significantly, our pipeline prioritizes maintaining style consistency across multiple objects within the scene, ensuring a cohesive and visually appealing result aligned with the desired aesthetic. The core strength of our pipeline lies in its ability to generate 3D scenes that are not only visually impressive but also exhibit features like photo-realistic, multi-view consistency, and diversity, demonstrating the versatility and adaptability of our model.

Controllable 3D indoor scene synthesis stands at the forefront of technological progress, offering various applications like gaming, film, and augmented/virtual reality. The capability to stylize and decouple objects within these scenarios is a crucial factor, providing an advanced level of control throughout the editing process. This control extends not just to manipulating geometric attributes like translation and scaling but also includes managing appearances, such as stylization. Current methods for scene stylization are limited to applying styles to the entire scene, without the ability to separate and customize individual objects. Addressing the intricacies of this challenge, we introduce a unique pipeline designed for synthesis 3D indoor scenes. Our approach involves strategically placing objects within the scene, utilizing information from professionally designed bounding boxes. Significantly, our pipeline prioritizes maintaining style consistency across multiple objects within the scene, ensuring a cohesive and visually appealing result aligned with the desired aesthetic. The core strength of our pipeline lies in its ability to generate 3D scenes that are not only visually impressive but also exhibit features like photo-realistic, multi-view consistency, and diversity, demonstrating the versatility and adaptability of our model.

Event-based structured light (SL) systems have attracted increasing attention for their potential in high-performance 3D measurement. Despite the inherent HDR capability of event cameras, reflective and absorptive surfaces still cause event cluttering and absence, which produce overexposed and underexposed regions that degrade the reconstruction quality. In this work, we present the first HDR 3D measurement framework specifically designed for event-based SL systems. First, we introduce a multi-contrast HDR coding strategy that facilitates imaging of areas with different reflectance. Second, to alleviate inter-frame interference caused by overexposed and underexposed areas, we propose a universal confidence-driven stereo matching strategy. Specifically, we estimate a confidence map as the fusion weight for features via an energy-guided confidence estimation. Further, we propose the confidence propagation volume, an innovative cost volume that offers both effective suppression of inter-frame interference and strong representation capability. Third, we contribute an event-based SL simulator and propose the first event-based HDR SL dataset. We also collect a real-world benchmarking dataset with ground truth. We validate the effectiveness of our method with the proposed confidence-driven strategy on both synthetic and real-world datasets. Experimental results demonstrate that our proposed HDR framework enables accurate 3D measurement even under extreme conditions.

From camera to screen, researchers have developed a well-established system for capturing and reproducing the color experience of human eyes. In this study, we aim to upgrade this process by transiting from conventional RGB to multi-spectral image (MSI) color reproduction. While MSI offers evident advantages in color matching, we find out it is not trivial to make good use of more spectral information for color constancy. Therefore, we present a regularized color reproduction system that incorporates a spectral prior-guided optimization strategy to establish a sensor-optimized RGB projection for color matching, along with a learning-based chromatic adaptation model for color constancy. Specifically, we define the RGB projection through an end-to-end optimization under the guidance of sensor spectral sensitivities. Subsequently, we devise a chromatic adaptation neural network that estimates the scene illuminance and an illuminance-adaptive matrix for auto white balancing and dynamic color correction, respectively. Comprehensive experiments show the superiority of our system compared to alternative solutions.

Visual images corrupted by various types and levels of degradations are commonly encountered in practical image compression. However, most existing image compression methods are tailored for clean images, therefore struggling to achieve satisfying results on these images. Joint compression and restoration methods typically focus on a single type of degradation and fail to address a variety of degradations in practice. To this end, we propose a unified framework for all-in-one image compression and restoration, which incorporates the image restoration capability against various degradations into the process of image compression. The key challenges involve distinguishing authentic image content from degradations, and flexibly eliminating various degradations without prior knowledge. Specifically, the proposed framework approaches these challenges from two perspectives: i.e., content information aggregation, and degradation representation aggregation. Extensive experiments demonstrate the following merits of our model: 1) superior rate-distortion (RD) performance on various degraded inputs while preserving the performance on clean data; 2) strong generalization ability to real-world and unseen scenarios; 3) higher computing efficiency over compared methods. Our code is available at https://github.com/ZeldaM1/All-in-one.

Recent advancements in light field super-resolution (SR) have yielded impressive results. In practice, however, many existing methods are limited by assuming fixed degradation models , such as bicubic downsampling, which hinders their robustness in real-world scenarios with complex degradations. To address this limitation, we present LF-DEST, an effective blind L ight F ield SR method that incorporates explicit D egradation Est imation to handle various degradation types. LF-DEST consists of two primary components: degradation estimation and light field restoration. The former concurrently estimates blur kernels and noise maps from low-resolution degraded light fields, while the latter generates super-resolved light fields based on the estimated degradations. Notably, we introduce a modulated and selective fusion module that intelligently combines degradation representations with image information, effectively handling diverse degradation types. We conduct extensive experiments on benchmark datasets, demonstrating that LF-DEST achieves superior performance across various degradation scenarios in light field SR. The implementation code is available at https://github.com/zeyuxiao1997/LF-DEST .

Image resampling is a basic technique that is widely employed in daily applications, such as camera photo editing. Recent deep neural networks (DNNs) have made impressive progress in performance by introducing learned data priors. Still, these methods are not the perfect substitute for interpolation, due to the drawbacks in efficiency and versatility. In this work, we propose a novel method of Learning Resampling Function (termed LeRF), which takes advantage of both the structural priors learned by DNNs and the locally continuous assumption of interpolation. Specifically, LeRF assigns spatially varying resampling functions to input image pixels and learns to predict the hyper-parameters that determine the shapes of these resampling functions with a neural network. Based on the formulation of LeRF, we develop a family of models, including both efficiency-orientated and performance-orientated ones. To achieve interpolation-level efficiency, we adopt look-up tables (LUTs) to accelerate the inference of the learned neural network. Furthermore, we design a directional ensemble strategy and edge-sensitive indexing patterns to better capture local structures. On the other hand, to obtain DNN-level performance, we propose an extension of LeRF to enable it in cooperation with pre-trained upsampling models for cascaded resampling. Extensive experiments show that the efficiency-orientated version of LeRF runs as fast as interpolation, generalizes well to arbitrary transformations, and outperforms interpolation significantly, e.g., up to 3 dB PSNR gain over Bicubic for $\times 2$×2 upsampling on Manga109. Besides, the performance-orientated version of LeRF reaches comparable performance with existing DNNs at much higher efficiency, e.g., less than 25% running time on a desktop GPU.

The fusion of low-spatial-resolution hyperspectral image (LR-HSI) with high-spatial-resolution multispectral image (HR-MSI) has become an effective way to obtain the high-spatial-resolution hyperspectral image (HR-HSI). Currently, learning-based methods have emerged as the mainstream solution in this field. However, these methods typically rely on predefined or simplified degradation models during fusion training, resulting in inaccurate supervision of the fusion networks. Meanwhile, most methods overlook the degradation characteristics in designing the fusion networks, leading to a mismatch between the degradation and fusion processes. These limitations ultimately result in unsatisfactory fusion performance on real data. To enhance the practicality of learning-based methods, accurate degradation modeling and effective network design have become the critical priorities. We observe that, in practical scenarios, the degree of pixel degradation varies across different positions due to the unforeseen factors such as illumination variations and imaging system fluctuations. Considering this, we propose a non-uniform degradation model (NUD), which introduces non-uniformity into the degradation processes of LR-HSI and HR-MSI. In addition, we emphasize that the essence of fusion is to reverse the degradation process. Therefore, to align with the non-uniform degradation process, the fusion process should exhibit similar positional specificity. For this purpose, we propose a position-aware fusion network (PAF), which employs positional encoding to endow the fusion process with the position-aware attribute. Experimental results show that our proposed methods provide an effective solution for HSI fusion in practical scenarios.

Non-line-of-sight (NLOS) imaging aims to reconstruct scenes hidden from direct view and has broad applications in robotic vision, rescue operations, autonomous driving, and remote sensing. However, most existing methods rely on densely sampled transients from large, continuous relay surfaces, which limits their practicality in real-world scenarios with aperture constraints. To address this limitation, we propose an unsupervised zero-shot framework tailored for confocal NLOS imaging with aperture-limited relay surfaces. Our method leverages latent diffusion models to recover fully-sampled transients from undersampled versions by enforcing measurement consistency during the sampling process. To further improve recovered transient quality, we introduce a progressive recovery strategy that incrementally recovers missing transient values, effectively mitigating the impact of severe aperture limitations. In addition, to suppress error propagation during recovery, we develop a backpropagation-based error correction reconstruction algorithm that refines intermediate recovered transients by enforcing sparsity regularization in the voxel domain, enabling high-fidelity final reconstructions. Extensive experiments on both simulated and real-world datasets validate the robustness and generalization capability of our method across diverse aperture-limited relay surfaces. Notably, our method follows a zero-shot paradigm, requiring only a single pretraining stage without paired data or pattern-specific retraining, which makes it a more practical and generalizable framework for NLOS imaging.

Video restoration from low-resolution and low-frame-rate blurry sources remains challenging due to insufficient data priors. In this paper, we propose BVSR-EvD, leveraging event cameras and diffusion models to boost blurry video space-time super-resolution. Specifically, we identify three distinct data priors from event-video dual modalities: motion prior from events, content prior from videos, and physical prior from their integration, contributing to temporal stability, content preservation, and detail enhancement respectively. To effectively utilize these data priors, BVSR-EvD creates the Trident Diffusion Model (Trident-DM), which decomposes each denoising step into trident decoupling and adaptive self-composition stages. The former employs single-modal and dual-modal meta-networks to extract the three unique data priors, while the latter dynamically integrates them through learned prior-aware weight maps. BVSR-EvD achieves up to $\times 8$ spatial super-resolution and $\times 64$ temporal super-resolution from blurry videos, surpassing existing methods on public video datasets.

Proposal-free instance segmentation methods have significantly advanced the field of biological image analysis. Recently, the Segment Anything Model (SAM) has shown an extraordinary ability to handle challenging instance boundaries. However, directly applying SAM to biological images that contain instances with complex morphologies and dense distributions fails to yield satisfactory results. In this work, we propose BioSAM, a new biological instance segmentation framework generating SAM prompts from a superpixel graph. Specifically, to avoid over-merging, we first generate sufficient superpixels as graph nodes and construct an initialized graph. We then generate initial prompts from each superpixel and aggregate them through a graph neural network (GNN) by predicting the relationship of superpixels to avoid over-segmentation. We employ the SAM encoder embeddings and the SAM-assisted superpixel similarity as new features for the graph to enhance its discrimination capability. With the graph-based prompt aggregation, we utilize the aggregated prompts in SAM to refine the segmentation and generate more accurate instance boundaries. Comprehensive experiments on four representative biological datasets demonstrate that our proposed method outperforms state-of-the-art methods.

Electron microscopy (EM) image denoising is critical for visualization and subsequent analysis. Despite the remarkable achievements of deep learning-based non-blind denoising methods, their performance drops significantly when domain shifts exist between the training and testing data. To address this issue, unpaired blind denoising methods have been proposed. However, these methods heavily rely on image-to-image translation and neglect the inherent characteristics of EM images, limiting their overall denoising performance. In this paper, we propose the first unsupervised domain adaptive EM image denoising method, which is grounded in the observation that EM images from similar samples share common content characteristics. Specifically, we first disentangle the content representations and the noise components from noisy images and establish a shared domain-agnostic content space via domain alignment to bridge the synthetic images (source domain) and the real images (target domain). To ensure precise domain alignment, we further incorporate domain regularization by enforcing that: the pseudo-noisy images, reconstructed using both content representations and noise components, accurately capture the characteristics of the noisy images from which the noise components originate, all while maintaining semantic consistency with the noisy images from which the content representations originate. To guarantee lossless representation decomposition and image reconstruction, we introduce disentanglement-reconstruction invertible networks. Finally, the reconstructed pseudo-noisy images, paired with their corresponding clean counterparts, serve as valuable training data for the denoising network. Extensive experiments on synthetic and real EM datasets demonstrate the superiority of our method in terms of image restoration quality and downstream neuron segmentation accuracy. Our code is publicly available at https://github.com/sydeng99/DADn.

Fine-grained entity recognition (FGER) attracts increasing attention in information extraction and many other natural language understanding applications. However, it is a quite challenging problem for a specific domain due to the lack of specific-domain labeled data. To address this challenge, recent advancements in language modeling such as generative pretrained transformer (GPT) offer promising alternatives. Since large language models (LLMs) can be used for various tasks, such as text generation, summarization, and information extraction without labeled data, we incorporated them into the FGER field. Nonetheless, when too many verbose labels are fed to LLMs simultaneously, LLMs occasionally generate content that diverges from user input, contradicts previously generated context, or misaligns with established world knowledge, also called the "hallucination" phenomenon. In this article, we propose a new method called FGER-GPT to address these issues. Our approach leverages multiple inference chains and incorporates a hierarchical strategy for recognizing fine-grained entities, resulting in a significant performance boost. Importantly, neither coarse-grained nor fine-grained entity annotations are used in our proposed approach, which avoids the heavy labor consumption of labeling. Extensive experiments conducted on widely used datasets have demonstrated that the proposed FGER-GPT achieves competitive performance compared to state-of-the-art approaches in low-resource scenarios, highlighting its feasibility for real-world applications.

Recent research has explored leveraging event cameras, known for their prowess in capturing scenes with nonuniform motion, for video deraining, leading to performance improvements. However, the existing event-based method still faces the challenge that the complex spatiotemporal distribution disrupts temporal information fusion and complicates feature separation. This article proposes a novel end-to-end learning framework for video deraining that effectively extracts the rich dynamic information provided by the event stream. Our framework incorporates two key modules: an event-aware motion detection (EAMD) module that adaptively aggregates multiframe motion information using event-driven masks and a pyramidal adaptive selection module that separates background and rain layers by leveraging contextual priors from both event and conventional camera data. To facilitate efficient training, we introduce a real-world dataset of synchronized rainy videos and event streams. Extensive evaluations on both synthetic and real-world datasets demonstrate the superiority of our proposed method compared to state-of-the-art approaches. The code is available at https://github.com/booker-max/EGVD.

Image registration plays a crucial role in histological image analysis, encompassing tasks like multi-modality fusion and disease grading. Traditional registration methods optimize objective functions for each image pair, yielding reliable accuracy but demanding heavy inference burdens. Recently, learning-based registration methods utilize networks to learn the optimization process during training and apply a one-step forward process during testing. While these methods offer promising registration performance with reduced inference time, they remain sensitive to appearance variances and local structure changes commonly encountered in histological image registration scenarios. In this paper, for the first time, we propose a novel test-time adaptation method for histological image registration, aiming to improve the generalization ability of learning-based methods. Specifically, we design two operations, style guidance and shape guidance, for the test-time adaptation process. The former leverages style representations encoded by feature statistics to address the issue of appearance variances, while the latter incorporates shape representations encoded by HOG features to improve registration accuracy in regions with structural changes. Furthermore, we consider the continuity of the model during the test-time adaptation process. Different from the previous methods initialized by a given trained model, we introduce a smoothing strategy to leverage historical models for better generalization. We conduct experiments with several representative learning-based backbones on the public histological dataset, demonstrating the superior registration performance of our test-time adaptation method.

The current neuron reconstruction pipeline for electron microscopy (EM) data usually includes automatic image segmentation followed by extensive human expert proofreading. In this work, we aim to reduce human workload by predicting connectivity between over-segmented neuron pieces, taking both microscopy image and 3D morphology features into account, similar to human proofreading workflow. To this end, we first construct a dataset, named FlyTracing, that contains millions of pairwise connections of segments expanding the whole fly brain, which is three orders of magnitude larger than existing datasets for neuron segment connection. To learn sophisticated biological imaging features from the connectivity annotations, we propose a novel connectivity-aware contrastive learning method to generate dense volumetric EM image embedding. The learned embeddings can be easily incorporated with any point or voxel-based morphological representations for automatic neuron tracing. Extensive comparisons of different combination schemes of image and morphological representation in identifying split errors across the whole fly brain demonstrate the superiority of the proposed approach, especially for the locations that contain severe imaging artifacts, such as section missing and misalignment. The dataset and code are available at https://github.com/Levishery/Flywire-Neuron-Tracing.

Tumor synthesis enables the creation of artificial tumors in medical images, facilitating the training of AI models for tumor detection and segmentation. However, success in tumor synthesis hinges on creating visually realistic tumors that are generalizable across multiple organs and, furthermore, the resulting AI models being capable of detecting real tumors in images sourced from different domains (e.g., hospitals). This paper made a progressive stride toward generalizable tumor synthesis by leveraging a critical observation: early-stage tumors (< 2cm) tend to have similar imaging characteristics in computed tomography (CT), whether they originate in the liver, pancreas, or kidneys. We have ascertained that generative AI models, e.g., Diffusion Models, can create realistic tumors generalized to a range of organs even when trained on a limited number of tumor examples from only one organ. Moreover, we have shown that AI models trained on these synthetic tumors can be generalized to detect and segment real tumors from CT volumes, encompassing a broad spectrum of patient demographics, imaging protocols, and healthcare facilities.

Look-Up Table (LUT) has recently gained increasing at-tention for restoring High-Quality (HQ) images from Low-Quality (LQ) observations, thanks to its high computational efficiency achieved through a “space for time” strategy of caching learned LQ-HQ pairs. However, incorporating multiple LUTs for improved performance comes at the cost of a rapidly growing storage size, which is ultimately re-stricted by the allocatable on-device cache size. In this work, we propose a novel LUT compression framework to achieve a better trade-off between storage size and performance for LUT-based image restoration models. Based on the observation that most cached LQ image patches are dis-tributed along the diagonal of a LUT, we devise a Diagonal-First Compression (DFC) framework, where diagonal LQ-HQ pairs are preserved and carefully re-indexed to main-tain the representation capacity, while non-diagonal pairs are aggressively subsampled to save storage. Extensive ex-periments on representative image restoration tasks demon-strate that our DFC framework significantly reduces the storage size of LUT-based models (including our new de-sign) while maintaining their performance. For instance, DFC saves up to 90% of storage at a negligible performance drop for x 4 super-resolution. The source code is available on GitHub: https://github.com/leenas233IDFC.

In the realm of video object segmentation (VOS), the challenge of operating under low-light conditions persists, resulting in notably degraded image quality and compromised accuracy when comparing query and memory frames for similarity computation. Event cameras, characterized by their high dynamic range and ability to capture motion information of objects, offer promise in enhancing object visibility and aiding VOS methods under such low-light conditions. This paper introduces a pioneering framework tai-lored for low-light VOS, leveraging event camera data to elevate segmentation accuracy. Our approach hinges on two pivotal components: the Adaptive Cross-Modal Fusion (ACMF) module, aimed at extracting pertinent features while fusing image and event modalities to mitigate noise interference, and the Event-Guided Memory Matching (EGMM) module, designed to rectify the issue of in-accurate matching prevalent in low-light settings. Additionally, we present the creation of a synthetic LLE-DAVIS dataset and the curation of a real-world LLE-vas dataset, encompassing frames and events. Experimental evaluations corroborate the efficacy of our method across both datasets, affirming its effectiveness in low-light scenarios. The datasets are available at https://github.com/HebeiFast/EventLowLightVOS.

Accurate 3D neuron segmentation from electron mi-croscopy (EM) volumes is crucial for neuroscience re-search. However, the complex neuron morphology often leads to over-merge and over-segmentation results. Recent advancements utilize 3D CNNs to predict a 3D affinity map with improved accuracy but suffer from two challenges: high computational cost and limited input size, especially for practical deployment for large-scale EM volumes. To address these challenges, we propose a novel method to leverage lightweight 2D CNNs for efficient neuron segmen-tation. Our method employs a 2D Y-shape network to generate two embedding maps from adjacent 2D sections, which are then converted into an affinity map by measuring their embedding distance. While the 2D network better captures pixel dependencies inside sections with larger in-put sizes, it overlooks inter-section dependencies. To over-come this, we introduce a cross-dimension affinity distillation (CAD) strategy that transfers inter-section dependency knowledge from a 3D teacher network to the 2D student network by ensuring consistency between their output affin-ity maps. Additionally, we design a feature grafting in-teraction (FGI) module to enhance knowledge transfer by grafting embedding maps from the 2D student onto those from the 3D teacher. Extensive experiments on multiple EM neuron segmentation datasets, including a newly built one by ourselves, demonstrate that our method achieves supe-rior performance over state-of-the-art methods with only 1/20 inference latency. We release our code and dataset at https://github.com/liuxyll03/CAD.

As the mainstream technique for capturing images of biological specimens at nanometer resolution, electron microscopy (EM) is extremely time-consuming for scanning wide field-of-view (FOV) specimens. In this paper, we investigate a challenging task of large-factor EM image super-resolution (EMSR), which holds great promise for reducing scanning time, relaxing acquisition conditions, and expanding imaging FOV. By exploiting the repetitive structures and volumetric coherence of EM images, we propose the first generative learning-based framework for large-factor EMSR. Specifically, motivated by the predictability ofrepetitive structures and textures in EM images, we first learn a discrete codebook in the latent space to represent highresolution (HR) cell-specific priors and a latent vector indexer to map low-resolution (LR) EM images to their corresponding latent vectors in a generative manner. By incorporating the generative cell-specific priors from HR EM images through a multi-scale prior fusion module, we then deploy multi-image feature alignment and fusion to further exploit the inter-section coherence in the volumetric EM data. Extensive experiments demonstrate that our proposed framework outperforms advanced single-image and video super-resolution methods for 8× and 16× EMSR (i.e., with 64 times and 256 times less data acquired, respectively), achieving superior visual reconstruction quality and down-stream segmentation accuracy on benchmark EM datasets. Code is available at https://github.com/jtshou/GPEMSR.

In image restoration (IR), leveraging semantic priors from segmentation models has been a common approach to improve performance. The recent segment anything model (SAM) has emerged as a powerful tool for extracting advanced semantic priors to enhance IR tasks. However, the computational cost of SAM is prohibitive for IR, compared to existing smaller IR models. The incorporation of SAMfor extracting semantic priors considerably hampers the model inference efficiency. To address this issue, we propose a general framework to distill SAM's semantic knowledge to boost exiting IR models without interfering with their inference process. Specifically, our proposed framework consists of the semantic priors fusion (SPF) scheme and the semantic priors distillation (SPD) scheme. SPF fuses two kinds of information between the restored image predicted by the original IR model and the semantic mask predicted by SAM for the refined restored image. SPD leverages a self-distillation manner to distill the fused semantic priors to boost the performance of original IR models. Additionally, we design a semantic-guided relation (SGR) module for SPD, which ensures semantic feature representation space consistency to fully distill the priors. We demonstrate the effectiveness of our framework across multiple IR models and tasks, including deraining, deblurring, and denoising.

Electron microscopy (EM) images are notoriously challenging to segment due to their complex structures and lack of effective annotations. Fortunately, large-scale self-supervised pretraining offers a promising solution by allowing us to acquire prior knowledge of cell and subcellular tissue structures, which can significantly improve EM instance segmentation results. However, most existing pretraining methods fail to capture the crucial local information that is essential for EM images, instead focusing only on high-level semantic information. In this paper, we propose a novel pretraining framework that leverages multiscale visual representations to adapt to the complex structures of EM images. Our framework achieves instance-level alignment by maximizing the consistency between strongly and weakly augmented images, while also incorporating a cross-attention mechanism to match multiscale features and encode more low-level information into high-level semantics. Most importantly, our approach employs multi-task optimization on the feature pyramid, enabling multiscale pixel restoration and feature comparison. We extensively pretrain our method on four large-scale EM datasets and demonstrate significant gains on neuron and mitochondria segmentation tasks. Code is available at https://github.com/ydchen0806/MS-Con-EM-Seg.

Image aesthetics assessment (IAA) aims to estimate the aesthetics of images. Depending on the content of an image, diverse criteria need to be selected to assess its aesthetics. Existing works utilize pre-trained vision backbones based on content knowledge to learn image aesthetics. However, training those backbones is time-consuming and suffers from attention dispersion. Inspired by learnable queries in vision-language alignment, we propose the Image Aesthetics Assessment via Learnable Queries (IAA-LQ) approach. It adapts learnable queries to extract aesthetic features from pre-trained image features obtained from a frozen image encoder. Extensive experiments on real-world data demonstrate the advantages of IAA-LQ, beating the best state-of-the-art method by 2.2% and 2.1% in terms of SRCC and PLCC, respectively.

Understanding scene dynamics relies heavily on optical flow and scene flow. Most existing flow estimation methods use low-rate RGB images and point clouds, and match the frames geometrically. However, this approach faces challenges in real-world scenes with intricate motion, occlusion, and noise. To tackle this problem, we combine point clouds with events, which introduce dynamic inter-frame information. We propose a bi-stream neural network that jointly estimates optical flow and scene flow. The event branch extracts dynamic information and estimates optical flow, while the point branch captures scene structure and estimate scene flow. A Spatio-temporal Fusion Block is introduced to fuse the complementary information from points and events. Additionally, we adopt a result-level fusion strategy for direct refinement between the flow predictions of the two branches. We evaluate our model on the real-world datasets DSEC and MVSEC. The experimental results demonstrate superior performance compared to existing methods.

Image aesthetics assessment (IAA) is attracting wide interest with the prevalence of social media. The problem is challenging due to its subjective and ambiguous nature. Instead of directly extracting aesthetic features solely from the image, user comments associated with an image could potentially provide complementary knowledge that is useful for IAA. With existing large-scale pre-trained models demonstrating strong capabilities in extracting high-quality transferable visual and textual features, learnable queries are shown to be effective in extracting useful features from the pre-trained visual features. Therefore, in this paper, we propose MMLQ, which utilizes multi-modal learnable queries to extract aesthetics-related features from multi-modal pre-trained features. Extensive experimental results demonstrate that MMLQ achieves new state-of-the-art performance on multimodal IAA, beating previous methods by 7.7% and 8.3% in terms of SRCC and PLCC, respectively.

The advent of High Dynamic Range/Wide Color Gamut (HDR/WCG) display technology has made significant progress in providing exceptional richness and vibrancy for the human visual experience. However, the widespread adoption of HDR/WCG images is hindered by their substantial storage requirements, imposing significant bandwidth challenges during distribution. Besides, HDR/WCG images are often tone-mapped into Standard Dynamic Range (SDR) versions for compatibility, necessitating the usage of inverse Tone Mapping (iTM) techniques to reconstruct their original representation. In this work, we propose a meta-transfer learning framework for practical HDR/WCG media transmission by embedding image-wise metadata into their SDR counterparts for later iTM reconstruction. Specifically, we devise a meta-learning strategy to pre-train a lightweight multilayer perceptron (MLP) model that maps SDR pixels to HDR/WCG ones on an external dataset, resulting in a domain-wise iTM model. Subsequently, for the transfer learning process of each HDR/WCG image, we present a spatial-aware online mining mechanism to select challenging training pairs to adapt the meta-trained model to an image-wise iTM model. Finally, the adapted MLP, embedded as metadata, is transmitted alongside the SDR image, facilitating the reconstruction of the original image on HDR/WCG displays. We conduct extensive experiments and evaluate the proposed framework with diverse metrics. Compared with existing solutions, our framework shows superior performance in fidelity, minimal latency, and negligible overhead. The codes are available at https://github.com/pjliu3/MLP_iTM.

Event cameras are novel bio-inspired cameras that record asynchronous events with high temporal resolution and dynamic range. Leveraging the auxiliary temporal information recorded by event cameras holds great promise for the task of video super-resolution (VSR). However, existing event-guided VSR methods assume that the event and RGB cameras are strictly calibrated (e.g., pixel-level sensor designs in DAVIS 240/346). This assumption proves limiting in emerging high-resolution devices, such as dual-lens smartphones and unmanned aerial vehicles, where such precise calibration is typically unavailable. To unlock more event-guided application scenarios, we perform the task of asymmetric event-guided VSR for the first time, and we propose an Asymmetric Event-guided VSR Network (AsEVSRN) for this new task. AsEVSRN incorporates two specialized designs for leveraging the asymmetric event stream in VSR. Firstly, the content hallucination module dynamically enhances event and RGB information by exploiting their complementary nature, thereby adaptively boosting representational capacity. Secondly, the event-enhanced bidirectional recurrent cells align and propagate temporal features fused with features from content-hallucinated frames. Within the bidirectional recurrent cells, event-enhanced flow is employed to simultaneously utilize and fuse temporal information at both the feature and pixel levels. Comprehensive experimental results affirm that our method consistently generates superior quantitative and qualitative results.

In real-world photography, local motion blur often arises from the interplay between moving objects and stationary backgrounds during exposure. Existing deblurring methods face challenges in addressing local motion deblurring due to (i) the presence of arbitrary localized blurs and uncertain blur extents; (ii) the limited ability to accurately identify specific blurs resulting from ambiguous motion boundaries. These limitations often lead to suboptimal solutions when estimating blur maps and generating final deblurred images. To that end, we propose a novel method named Motion-Uncertainty-Guided Network (MUGNet), which harnesses a probabilistic representational model to explicitly address the intricacies stemming from motion uncertainties. Specifically, MUGNet consists of two key components, i.e., motion-uncertainty quantification (MUQ) module and motion-masked separable attention (M2SA) module, serving for complementary purposes. Concretely, MUQ aims to learn a conditional distribution for accurate and reliable blur map estimation, while the M2SA module is to enhance the representation of regions influenced by local motion blur and static background, which is achieved by promoting the establishment of extensive global interactions. We demonstrate the superiority of our MUGNet with extensive experiments. The code is publicly available at: https://github.com/zeyuxiao1997/MUGNet.

Dynamic reconstruction in confocal non-line-of-sight imaging encounters great challenges since the dense raster-scanning manner limits the practical frame rate. A fewer pioneer works reconstruct high-resolution volumes from the under-scanning transient measurements but overlook temporal consistency among transient frames. To fully exploit multi-frame information, we propose the first spatial-temporal Mamba (ST-Mamba) based method tailored for dynamic reconstruction of transient videos. Our method capitalizes on neighbouring transient frames to aggregate the target 3D hidden volume. Specifically, the interleaved features extracted from the input transient frames are fed to the proposed ST-Mamba blocks, which leverage the time-resolving causality in transient measurement. The cross ST-Mamba blocks are then devised to integrate the adjacent transient features. The target high-resolution transient frame is subsequently recovered by the transient spreading module. After transient fusion and recovery, a physical-based network is employed to reconstruct the hidden volume. To tackle the substantial noise inherent in transient videos, we propose a wave-based loss function to impose constraints within the phasor field. Besides, we introduce a new dataset, comprising synthetic videos for training and real-world videos for evaluation. Extensive experiments showcase the superior performance of our method on both synthetic data and real world data captured by different imaging setups. The code and data are available at https://github.com/Depth2World/Dynamic_NLOS.

Event cameras asynchronously measure brightness changes in a scene without motion blur or saturation, while frame cameras capture images with dense intensity and fine details at a fixed rate. The exclusive advantages of the two modalities make depth estimation from Stereo Asymmetric Frame-Event (SAFE) systems appealing. However, due to the inevitable information absence of one modality in certain challenging regions, existing stereo matching methods lose efficacy for asymmetric inputs from SAFE systems. In this paper, we propose a divide-and-conquer approach that decomposes depth estimation from SAFE systems into three sub-tasks, i.e., frame-event stereo matching, frame-based Structure-from-Motion (SfM), and event-based SfM. In this way, the above challenging regions are addressed by monocular SfM, which estimates robust depth with two views belonging to the same functioning modality. Moreover, we propose a dual sampling strategy to construct cost volumes with identical spatial locations and depth hypotheses for different sub-tasks, which enables sub-task fusion at the cost volume level. To tackle the occlusion issue raised by the sampling strategy, we further introduce a temporal fusion scheme to utilize long-term sequential inputs with multi-view information. Experimental results validate the superior performance of our method over existing solutions.

Video composition merges the foreground and background of different videos, presenting challenges due to variations in capture conditions (e.g., saturation, brightness, and contrast). Video harmonization is a vital process in achieving a realistic composite by seamlessly adjusting the foreground’s appearance to match the background. In this paper, we propose TSA2, a novel method for video harmonization that incorporates temporal segment adaptation and aggregation. TSA2divides the inharmonious input sequence into temporal segments, each corresponding to a different frame rate, allowing effective utilization of complementary information within each segment. The method includes the Temporal Segment Adaptation module, which learns and remaps the distribution difference between background and foreground regions, and the Temporal Segment Aggregation module, which emphasizes and aggregates cross-segment information through element-wise correlations. Experimental results demonstrate that TSA2outperforms advanced image and video harmonization methods quantitatively and qualitatively.

The widespread usage of high-definition screens on edge devices stimulates a strong demand for efficient image restoration algorithms. The way of caching deep learning models in a look-up table (LUT) is recently introduced to respond to this demand. However, the size of a single LUT grows exponentially with the increase of its indexing capacity, which restricts its receptive field and thus the performance. To overcome this intrinsic limitation of the single-LUT solution, we propose a universal method to construct multiple LUTs like a neural network, termed MuLUT. First, we devise novel complementary indexing patterns, as well as a general implementation for arbitrary patterns, to construct multiple LUTs in parallel. Second, we propose a re-indexing mechanism to enable hierarchical indexing between cascaded LUTs. Finally, we introduce channel indexing to allow cross-channel interaction, enabling LUTs to process color channels jointly. In these principled ways, the total size of MuLUT is linear to its indexing capacity, yielding a practical solution to obtain superior performance with the enlarged receptive field. We examine the advantage of MuLUT on various image restoration tasks, including super-resolution, demosaicing, denoising, and deblocking. MuLUT achieves a significant improvement over the single-LUT solution, e.g., up to 1.1 dB PSNR for super-resolution and up to 2.8 dB PSNR for grayscale denoising, while preserving its efficiency, which is 100× less in energy cost compared with lightweight deep neural networks.

Image denoising is a fundamental problem in computational photography, where achieving high perception with low distortion is highly demanding. Current methods either struggle with perceptual quality or suffer from significant distortion. Recently, the emerging diffusion model has achieved state-of-the-art performance in various tasks and demonstrates great potential for image denoising. However, stimulating diffusion models for image denoising is not straightforward and requires solving several critical problems. For one thing, the input inconsistency hinders the connection between diffusion models and image denoising. For another, the content inconsistency between the generated image and the desired denoised image introduces distortion. To tackle these problems, we present a novel strategy called the Diffusion Model for Image Denoising (DMID) by understanding and rethinking the diffusion model from a denoising perspective. Our DMID strategy includes an adaptive embedding method that embeds the noisy image into a pre-trained unconditional diffusion model and an adaptive ensembling method that reduces distortion in the denoised image. Our DMID strategy achieves state-of-the-art performance on both distortion-based and perception-based metrics, for both Gaussian and real-world image denoising.

Deep optics has been endeavoring to capture hyperspectral images of dynamic scenes, where the optical encoder plays an essential role in deciding the imaging performance. Our key insight is that the optical encoder of a deep optics system is expected to keep fabrication-friendliness and decoder-friendliness, to be faithfully realized in the implementation phase and fully interacted with the decoder in the design phase, respectively. In this paper, we propose the non-serial quantization-aware deep optics (NSQDO), which consists of the fabrication-friendly quantization-aware model (QAM) and the decoder-friendly non-serial manner (NSM). The QAM integrates the quantization process into the optimization and adaptively adjusts the physical height of each quantization level, reducing the deviation of the physical encoder from the numerical simulation through the awareness of and adaptation to the quantization operation of the DOE physical structure. The NSM bridges the encoder and the decoder with full interaction through bidirectional hint connections and flexibilize the connections with a gating mechanism, boosting the power of joint optimization in deep optics. The proposed NSQDO improves the fabrication-friendliness and decoder-friendliness of the encoder and develops the deep optics framework to be more practical and powerful. Extensive synthetic simulation and real hardware experiments demonstrate the superior performance of the proposed method.

Non-local self-similarity has been well exploited in the single image super-resolution task as an effective prior. However, due to the difficulty of modeling the 4D correspondence globally, the potential of the non-local prior is less revealed for light field (LF) super-resolution. Meanwhile, existing non-local models only utilize the global spatial correspondence, but largely neglect the global geometric correspondence. To address the aforementioned problems, we propose a Decoupled Selective Matching Network (DSMNet) for LF super-resolution, by designing a novel selective matching mechanism to flexibly extract non-local information from specific 4D positions in an LF. Such a mechanism matches the reference patch with several auxiliary patches dynamically searched from predefined windows, which promotes efficiency while improving performance compared to the existing non-local models. Specifically, our DSMNet decouples the whole LF into Sub-Aperture Images (SAIs) and Epipolar Plane Images (EPIs). For each SAI patch, we separately perform the selective matching inside the current SAI and cross different SAIs to exploit the global spatial correspondence efficiently. For each EPI patch, we separately perform the selective matching in EPIs of different orientations to embed robust LF geometric information into features by enhancing EPI textures, which exploits the global geometric correspondence in an efficient manner. Comprehensive experiments validate that DSMNet outperforms state-of-the-art LF super-resolution methods both quantitatively and qualitatively.

Biomedical videos require tremendous storage space and transmission bandwidth, so efficient coding methods are urgently required. Existing methods can be roughly divided into motion-based methods and wavelet-based methods. Motion-based methods use motion estimation designed for natural videos and independently optimize prediction, transform, and entropy coding modules. Wavelet-based methods treat the more redundant time dimension exactly the same as other spatial dimensions. They are both unable to completely remove the redundant spatial-temporal information in biomedical videos. In this paper, to address these problems, we build an end-to-end framework named DBVC with 3-D motion estimation, MV coding, 3-D motion compensation, and residual coding networks for efficient 3-D biomedical video coding. First, we propose a simple yet efficient 3-D motion estimation network to extract motion information. Specifically, we obtain the region with the most intense motion by a segmentation network and then perform unsupervised motion estimation exclusively on this region. After that, to encode and decode the estimated motion vectors, we apply a 3-D autoencoder-based MV coding network. Moreover, we use a lossless learnable wavelet transform for residual coding, which makes lossless coding possible. To the best of our knowledge, this is the first end-to-end video coding framework that supports both lossy and lossless coding, thus meeting the requirements of 3-D biomedical video coding. Extensive experiments demonstrate that our framework achieves state-of-the-art performance on both 3-D biological videos and 3-D medical videos.

Existing methods have demonstrated effective performance on a single degradation type. In practical applications, however, the degradation is often unknown, and the mismatch between the model and the degradation will result in a severe performance drop. In this paper, we propose an all-in-one image restoration network that tackles multiple degradations. Due to the heterogeneous nature of different types of degradations, it is difficult to process multiple degradations in a single network. To this end, we propose to learn a neural degradation representation (NDR) that captures the underlying characteristics of various degradations. The learned NDR adaptively decomposes different types of degradations, similar to a neural dictionary that represents basic degradation components. Subsequently, we develop a degradation query module and a degradation injection module to effectively approximate and utilize the specific degradation based on NDR, enabling the all-in-one restoration ability for multiple degradations. Moreover, we propose a bidirectional optimization strategy to effectively drive NDR to learn the degradation representation by optimizing the degradation and restoration processes alternately. Comprehensive experiments on representative types of degradations (including noise, haze, rain, and downsampling) demonstrate the effectiveness and generalizability of our method. Code is available at https://github.com/mdyao/NDR-Restore.

The size of image volumes in connectomics studies now reaches terabyte and often petabyte scales with a great diversity of appearance due to different sample preparation procedures. However, manual annotation of neuronal structures (e.g., synapses) in these huge image volumes is time-consuming, leading to limited labeled training data often smaller than 0.001% of the large-scale image volumes in application. Methods that can utilize in-domain labeled data and generalize to out-of-domain unlabeled data are in urgent need. Although many domain adaptation approaches are proposed to address such issues in the natural image domain, few of them have been evaluated on connectomics data due to a lack of domain adaptation benchmarks. Therefore, to enable developments of domain adaptive synapse detection methods for large-scale connectomics applications, we annotated 14 image volumes from a biologically diverse set of Megaphragma viggianii brain regions originating from three different whole-brain datasets and organized the WASPSYN challenge at ISBI 2023. The annotations include coordinates of pre-synapses and post-synapses in the 3D space, together with their one-to-many connectivity information. This paper describes the dataset, the tasks, the proposed baseline, the evaluation method, and the results of the challenge. Limitations of the challenge and the impact on neuroscience research are also discussed. The challenge is and will continue to be available at https://codalab.lisn.upsaclay.fr/competitions/9169. Successful algorithms that emerge from our challenge may potentially revolutionize real-world connectomics research and further the cause that aims to unravel the complexity of brain structure and function.

Portrait retouching aims to improve the aesthetic quality of input portrait photos and especially requires human-region priority. The deep learning-based methods largely elevate the retouching efficiency and provide promising retouched results. However, existing portrait retouching methods focus on automatic retouching, which treats all human-regions equally and ignores users' preferences for specific individuals, thus suffering from limited flexibility in interactive scenarios. In this work, we emphasize the importance of users' intents and explore the interactive portrait retouching task. Specifically, we propose a region-aware retouching framework with two branches: an automatic branch and an interactive branch. The automatic branch involves an encoding-decoding process, which searches region candidates and performs automatic region-aware retouching without user guidance. The interactive branch encodes sparse user guidance into a priority condition vector and modulates latent features with a region selection module to further emphasize the user-specified regions. Experimental results show that our interactive branch effectively captures users' intents and generalizes well to unseen scenes with sparse user guidance, while our automatic branch also outperforms the state-of-the-art retouching methods due to improved region-awareness.

The analysis of neuronal morphological data is essential to investigate the neuronal properties and brain mechanisms. The complex morphologies, absence of annotations, and sheer volume of these data pose significant challenges in neuronal morphological analysis, such as identifying neuron types and large-scale neuron retrieval, all of which require accurate measuring and efficient matching algorithms. Recently, many studies have been conducted to describe neuronal morphologies quantitatively using predefined measurements. However, hand-crafted features are usually inadequate for distinguishing fine-grained differences among massive neurons. In this article, we propose a novel morphology-aware contrastive graph neural network (MACGNN) for unsupervised neuronal morphological representation learning. To improve the retrieval efficiency in large-scale neuronal morphological datasets, we further propose Hash-MACGNN by introducing an improved deep hash algorithm to train the network end-to-end to learn binary hash representations of neurons. We conduct extensive experiments on the largest dataset, NeuroMorpho, which contains more than 100 000 neurons. The experimental results demonstrate the effectiveness and superiority of our MACGNN and Hash-MACGNN for large-scale neuronal morphological analysis.

Reflection from glasses is ubiquitous in daily life, but it is usually undesirable in photographs. To remove these unwanted noises, existing methods utilize either correlative auxiliary information or handcrafted priors to constrain this ill-posed problem. However, due to their limited capability to describe the properties of reflections, these methods are unable to handle strong and complex reflection scenes. In this article, we propose a hue guidance network (HGNet) with two branches for single image reflection removal (SIRR) by integrating image information and corresponding hue information. The complementarity between image information and hue information has not been noticed. The key to this idea is that we found that hue information can describe reflections well and thus can be used as a superior constraint for the specific SIRR task. Accordingly, the first branch extracts the salient reflection features by directly estimating the hue map. The second branch leverages these effective features, which can help locate salient reflection regions to obtain a high-quality restored image. Furthermore, we design a new cyclic hue loss to provide a more accurate optimization direction for the network training. Experiments substantiate the superiority of our network, especially its excellent generalization ability to various reflection scenes, as compared with state-of-the-arts both qualitatively and quantitatively. Source codes are available at https://github.com/zhuyr97/HGRR.

Depth estimation from indoor panoramas is challenging due to the equirectangular distortions of panoramas and inaccurate matching. In this paper, we propose a practical framework to improve the accuracy and efficiency of depth estimation from multi-view indoor panoramic images with the Neural Radiance Field technology. Specifically, we develop two networks to implicitly learn the Signed Distance Function for depth measurements and the radiance field from panoramas. We also introduce a novel spherical position embedding scheme to achieve high accuracy. For better convergence, we propose an initialization method for the network weights based on the Manhattan World Assumption. Furthermore, we devise a geometric consistency loss, leveraging the surface normal, to further refine the depth estimation. The experimental results demonstrate that our proposed method outperforms state-of-the-art works by a large margin in both quantitative and qualitative evaluations. Our source code is available at https://github.com/WJ-Chang-42/IndoorPanoDepth.

Exposure correction task aims to correct the underexposure and its adverse overexposure images to the normal exposure in a single network. As well recognized, the optimization flow is the opposite. Despite great advancement, existing exposure correction methods are usually trained with a mini-batch of both underexposure and overexposure mixed samples and have not explored the relationship between them to solve the optimization inconsistency. In this paper, we introduce a new perspective to conjunct their optimization processes by correlating and constraining the relationship of correction procedure in a mini-batch. The core designs of our framework consist of two steps: 1) formulating the exposure relationship of samples across the batch dimension via a context-irrelevant pretext task. 2) delivering the above sample relationship design as the regularization term within the loss function to promote optimization consistency. The proposed sample relationship design as a general term can be easily integrated into existing exposure correction methods without any computational burden in inference time. Extensive experiments over multiple representative exposure correction benchmarks demonstrate consistent performance gains by introducing our sample relationship design.

Existing semantic segmentation methods improve generalization capability, by regularizing various images to a canonical feature space. While this process contributes to generalization, it weakens the representation inevitably. In contrast to existing methods, we instead utilize the difference between images to build a better representation space, where the distinct style features are extracted and stored as the bases of representation. Then, the generalization to unseen image styles is achieved by projecting features to this known space. Specifically, we realize the style projection as a weighted combination of stored bases, where the similarity distances are adopted as the weighting factors. Based on the same concept, we extend this process to the decision part of model and promote the generalization of semantic prediction. By measuring the similarity distances to semantic bases (i.e., prototypes), we replace the common deterministic prediction with semantic clustering. Comprehensive experiments demonstrate the advantage of proposed method to the state of the art, up to 3.6% mIoU improvement in average on unseen scenarios. Code and models are available at https://gitee.com/mindspore/models/tree/master/research/cv/SPC-Net.

In this paper, we propose an efficient event-based motion estimation framework for various motion models. Different from previous works, we design a progressive event-to-map alignment scheme and utilize the spatio-temporal correlations to align events. In detail, we progressively align sampled events in an event batch to the time-surface map and obtain the updated motion model by minimizing a novel time-surface loss. In addition, a dynamic batch size strategy is applied to adaptively adjust the batch size so that all events in the batch are consistent with the current motion model. Our framework has three advantages: a) the progressive scheme refines motion parameters iteratively, achieving accurate motion estimation; b) within one iteration, only a small portion of events are involved in optimization, which greatly reduces the total runtime; c) the dynamic batch size strategy ensures that the constant velocity assumption always holds. We conduct comprehensive experiments to evaluate our framework on challenging high-speed scenes with three motion models: rotational, homography, and 6-DOF models. Experimental results demonstrate that our framework achieves state-of-the-art estimation accuracy and efficiency. The code is available at https://github.com/huangxueyan/PEME.

Image resampling is a basic technique that is widely employed in daily applications. Existing deep neural networks (DNNs) have made impressive progress in resampling performance. Yet these methods are still not the perfect substitute for interpolation, due to the issues of efficiency and continuous resampling. In this work, we propose a novel method of Learning Resampling Function (termed LeRF), which takes advantage of both the structural priors learned by DNNs and the locally continuous assumption of interpolation methods. Specifically, LeRF assigns spatially-varying steerable resampling functions to input image pixels and learns to predict the hyper-parameters that determine the orientations of these resampling functions with a neural network. To achieve highly efficient inference, we adopt look-up tables (LUTs) to accelerate the inference of the learned neural network. Furthermore, we design a directional ensemble strategy and edge-sensitive indexing patterns to better capture local structures. Extensive experiments show that our method runs as fast as interpolation, generalizes well to arbitrary transformations, and outperforms interpolation significantly, e.g., up to 3dB PSNR gain over bicubic for x 2 upsampling on Manga109.

Time-resolved non-line-of-sight (NLOS) imaging is based on the multi-bounce indirect reflections from the hidden objects for 3D sensing. Reconstruction from NLOS measurements remains challenging especially for complicated scenes. To boost the performance, we present NLOST, the first transformer-based neural network for NLOS reconstruction. Specifically, after extracting the shallow features with the assistance of physics-based priors, we design two spatial-temporal self attention encoders to explore both local and global correlations within 3D NLOS data by splitting or downsampling the features into different scales, respectively. Then, we design a spatial-temporal cross attention decoder to integrate local and global features in the token space of transformer, resulting in deep features with high representation capabilities. Finally, deep and shallow features are fused to reconstruct the 3D volume of hidden scenes. Extensive experimental results demonstrate the superior performance of the proposed method over existing solutions on both synthetic data and real-world data captured by different NLOS imaging systems.

Neuron reconstruction in a full adult fly brain from high-resolution electron microscopy (EM) data is regarded as a cornerstone for neuroscientists to explore how neurons inspire intelligence. As the central part of neurons, somas in the full brain indicate the origin of neurogenesis and neural functions. However, due to the absence of EM datasets specifically annotated for somas, existing deep learning-based neuron reconstruction methods cannot directly provide accurate soma distribution and morphology. Moreover, full brain neuron reconstruction remains extremely time-consuming due to the unprecedentedly large size of EM data. In this paper, we develop an efficient soma reconstruction method for obtaining accurate soma distribution and morphology information in a full adult fly brain. To this end, we first make a high-resolution EM dataset with fine-grained 3D manual annotations on somas. Relying on this dataset, we propose an efficient, two-stage deep learning algorithm for predicting accurate locations and boundaries of 3D soma instances. Further, we deploy a parallelized, high-throughput data processing pipeline for executing the above algorithm on the full brain. Finally, we provide quantitative and qualitative benchmark comparisons on the testset to validate the superiority of the proposed method, as well as preliminary statistics of the reconstructed somas in the full adult fly brain from the biological perspective. We release our code and dataset at https://github.com/liuxy1103/EMADS.

Vision-language models (VLMs) can effectively transfer to various vision tasks via prompt learning. Real-world scenarios often require adapting a model to multiple similar yet distinct tasks. Existing methods focus on learning a specific prompt for each task, limiting the ability to exploit potentially shared information from other tasks. Naively training a task-shared prompt using a combination of all tasks ignores fine-grained task correlations. Significant discrepancies across tasks could cause negative transferring. Considering this, we present Hierarchical Prompt (HiPro) learning, a simple and effective method for jointly adapting a pre-trained VLM to multiple downstream tasks. Our method quantifies inter-task affinity and subsequently constructs a hierarchical task tree. Task-shared prompts learned by internal nodes explore the information within the corresponding task group, while task-individual prompts learned by leaf nodes obtain fine-grained information targeted at each task. The combination of hierarchical prompts provides high-quality content of different granularity. We evaluate HiPro on four multi-task learning datasets. The results demonstrate the effectiveness of our method.

Camouflaged Instance Segmentation (CIS) aims at predicting the instance-level masks of camouflaged objects, which are usually the animals in the wild adapting their appearance to match the surroundings. Previous instance segmentation methods perform poorly on this task as they are easily disturbed by the deceptive camouflage. To address these challenges, we propose a novel De-camouflaging Network (DCNet) including a pixel-level camouflage decoupling module and an instance-level camouflage suppression module. The proposed DCNet enjoys several merits. First, the pixel-level camouflage decoupling module can extract camouflage characteristics based on the Fourier transformation. Then a difference attention mechanism is proposed to eliminate the camouflage characteristics while reserving target object characteristics in the pixel feature. Second, the instance-level camouflage suppression module can aggregate rich instance information from pixels by use of instance prototypes. To mitigate the effect of background noise during segmentation, we introduce some reliable reference points to build a more robust similarity measurement. With the aid of these two modules, our DCNet can effectively model de-camouflaging and achieve accurate segmentation for camouflaged instances. Extensive experimental results on two benchmarks demonstrate that our DCNet performs favorably against state-of-the-art CIS methods, e.g., with more than 5% performance gains on COD10K and NC4K datasets in average precision.

Automatic mitochondria segmentation enjoys great popularity with the development of deep learning. However, existing methods rely heavily on the labor-intensive manual gathering by experienced domain experts. And naively applying semi-supervised segmentation methods in the natural image field to mitigate the labeling cost is undesirable. In this work, we analyze the gap between mitochondrial images and natural images and rethink how to achieve effective semi-supervised mitochondria segmentation, from the perspective of reliable prototype-level supervision. We propose a novel end-to-end dual-reliable (DualRel) network, including a reliable pixel aggregation module and a reliable prototype selection module. The proposed DualRel enjoys several merits. First, to learn the prototypes well without any explicit supervision, we carefully design the referential correlation to rectify the direct pairwise correlation. Second, the reliable prototype selection module is responsible for further evaluating the reliability of prototypes in constructing prototype-level consistency regularization. Extensive experimental results on three challenging benchmarks demonstrate that our method performs favorably against state-of-the-art semi-supervised segmentation methods. Importantly, with extremely few samples used for training, DualRel is also on par with current state-of-the-art fully supervised methods.

Restoring sharp high frame-rate videos from low frame-rate blurry videos is a challenging problem. Existing blurry frame interpolation methods assume a predefined and known exposure time, which suffer from severe performance drop when applied to videos captured in the wild. In this paper, we study the problem of blurry frame interpolation under blind exposure with the assistance of an event camera. The high temporal resolution of the event camera is beneficial to obtain the exposure prior that is lost during the imaging process. Besides, sharp frames can be restored using event streams and blurry frames relying on the mutual constraint among them. Therefore, we first propose an exposure estimation strategy guided by event streams to estimate the lost exposure prior, transforming the blind exposure problem well-posed. Second, we propose to model the mutual constraint with a temporal-exposure control strategy through iterative residual learning. Our blurry frame interpolation method achieves a distinct performance boost over existing methods on both synthetic and self-collected real- world datasets under blind exposure.

Data augmentation (DA) is an efficient strategy for improving the performance of deep neural networks. Recent DA strategies have demonstrated utility in single image super-resolution (SR). Little research has, however, focused on the DA strategy for light field SR, in which multi-view information utilization is required. For the first time in light field SR, we propose a potent DA strategy called CutMIB to improve the performance of existing light field SR networks while keeping their structures unchanged. Specifically, Cut-MIB first cuts low-resolution (LR) patches from each view at the same location. Then CutMIB blends all LR patches to generate the blended patch and finally pastes the blended patch to the corresponding regions of high-resolution light field views, and vice versa. By doing so, CutMIB enables light field SR networks to learn from implicit geometric information during the training stage. Experimental results demonstrate that CutMIB can improve the reconstruction performance and the angular consistency of existing light field SR networks. We further verify the effectiveness of CutMIB on real-world light field SR and light field denoising. The implementation code is available at https://github.com/zeyuxiao1997/CutMIB.

In many computer vision applications (e.g., robotics and autonomous driving), high dynamic range (HDR) data is necessary for object detection algorithms to handle a variety of lighting conditions, such as strong glare. In this paper, we aim to achieve object detection on RAW sensor data, which naturally saves the HDR information from image sensors without extra equipment costs. We build a novel RAW sensor dataset, named ROD, for Deep Neural Networks (DNNs)-based object detection algorithms to be applied to HDR data. The ROD dataset contains a large amount of annotated instances of day and night driving scenes in 24-bit dynamic range. Based on the dataset, we first investigate the impact of dynamic range for DNNs-based detectors and demonstrate the importance of dynamic range adjustment for detection on RAW sensor data. Then, we propose a simple and effective adjustment method for object detection on HDR RAW sensor data, which is image adaptive and jointly optimized with the downstream detector in an end-to-end scheme. Extensive experiments demonstrate that the performance of detection on RAW sensor data is significantly superior to standard dynamic range (SDR) data in different situations. Moreover, we analyze the influence of texture information and pixel distribution of input data on the performance of the DNNs-based detector. Code and dataset will be available at https://gitee.com//mindspore/models/tree/master/research/cv/RAOD.

The asymmetric dual-lens configuration is commonly available on mobile devices nowadays, which naturally stores a pair of wide-angle and telephoto images of the same scene to support realistic super-resolution (SR). Even on the same device, however, the degradation for modeling realistic SR is image-specific due to the unknown acquisition process (e.g., tiny camera motion). In this paper, we propose a zero-shot solution for dual-lens SR (ZeDuSR), where only the dual-lens pair at test time is used to learn an image-specific SR model. As such, ZeDuSR adapts itself to the current scene without using external training data, and thus gets rid of generalization difficulty. However, there are two major challenges to achieving this goal: 1) dual-lens alignment while keeping the realistic degradation, and 2) effective usage of highly limited training data. To overcome these two challenges, we propose a degradation-invariant alignment method and a degradation-aware training strategy to fully exploit the information within a single dual-lens pair. Extensive experiments validate the superiority of Ze-DuSR over existing solutions on both synthesized and real-world dual-lens datasets. The implementation code is available at https://github.com/XrKang/ZeDuSR.

Sparse instance-level supervision has recently been explored to address insufficient annotation in biomedical instance segmentation, which is easier to annotate crowded instances and better preserves instance completeness for 3D volumetric datasets compared to common semi-supervision. In this paper, we propose a sparsely supervised biomedical instance segmentation framework via cross-representation affinity consistency regularization. Specifically, we adopt two individual networks to enforce the perturbation consistency between an explicit affinity map and an implicit affinity map to capture both feature-level instance discrimination and pixel-level instance boundary structure. We then select the highly confident region of each affinity map as the pseudo label to supervise the other one for affinity consistency learning. To obtain the highly confident region, we propose a pseudo-label noise filtering scheme by integrating two entropy-based decision strategies. Extensive experiments on four biomedical datasets with sparse instance annotations show the state-of-the-art performance of our proposed framework. For the first time, we demonstrate the superiority of sparse instance-level supervision on 3D volumetric datasets, compared to common semi-supervision under the same annotation cost. Code is available at https://github.com/liuxy1103/CRAC.

Mitochondria, as tiny structures within the cell, are of significant importance in studying cell functions for biological and clinical analysis. And exploring how to automatically segment mitochondria in electron microscopy (EM) images has attracted increasing attention. However, most of existing methods struggle to adapt to different scales and appearances of the input due to the inherent limitations of the traditional CNN architecture. To mitigate these limitations, we propose a novel adaptive template transformer (ATFormer) for mitochondria segmentation. The proposed ATFormer model enjoys several merits. First, the designed structural template learning module can acquire appearance-adaptive templates of background, foreground and contour to sense the characteristics of different shapes of mitochondria. And we further adopt an optimal transport algorithm to enlarge the discrepancy among diverse templates to activate corresponding regions fully. Second, we introduce a hierarchical attention learning mechanism to absorb multi-level information for templates to be adaptive scale-aware classifiers for dense prediction. Extensive experimental results on three challenging benchmarks including MitoEM, Lucchi and NucMM-Z datasets demonstrate that our ATFormer performs favorably against state-of-the-art mitochondria segmentation methods.

Event cameras are a type of novel neuromorphic sensor that has been gaining increasing attention. Existing event-based backbones mainly rely on image-based designs to extract spatial information within the image transformed from events, overlooking important event properties like time and polarity. To address this issue, we propose a novel Group-based vision Transformer backbone for Event-based vision, called Group Event Transformer (GET), which decouples temporal-polarity information from spatial information throughout the feature extraction process. Specifically, we first propose a new event representation for GET, named Group Token, which groups asynchronous events based on their timestamps and polarities. Then, GET applies the Event Dual Self-Attention block, and Group Token Aggregation module to facilitate effective feature communication and integration in both the spatial and temporal-polarity domains. After that, GET can be integrated with different downstream tasks by connecting it with various heads. We evaluate our method on four event-based classification datasets (Cifar10-DVS, N-MNIST, N-CARS, and DVS128Gesture) and two event-based object detection datasets (1Mpx and Gen1), and the results demonstrate that GET outperforms other state-of-the-art methods. The code is available at https://github.com/Peterande/GET-Group-Event-Transformer.

Current unsupervised video deraining methods are inefficient in modeling the intricate spatio-temporal properties of rain, which leads to unsatisfactory results. In this paper, we propose a novel approach by integrating a bio-inspired event camera into the unsupervised video deraining pipeline, which enables us to capture high temporal resolution information and model complex rain characteristics. Specifically, we first design an end-to-end learning-based network consisting of two modules, the asymmetric separation module and the cross-modal fusion module. The two modules are responsible for segregating the features of the rain-background layer, and for positive enhancement and negative suppression from a cross-modal perspective, respectively. Second, to regularize the network training, we elaborately design a cross-modal contrastive learning method that leverages the complementary information from event cameras, exploring the mutual exclusion and similarity of rain-background layers in different domains. This encourages the deraining network to focus on the distinctive characteristics of each layer and learn a more discriminative representation. Moreover, we construct the first real-world dataset comprising rainy videos and events using a hybrid imaging system. Extensive experiments demonstrate the superior performance of our method on both synthetic and real-world datasets.

Lightness adaptation is vital to the success of image processing to avoid unexpected visual deterioration, which covers multiple aspects, e.g., low-light image enhancement, image retouching, and inverse tone mapping. Existing methods typically work well on their trained lightness conditions but perform poorly in unknown ones due to their limited generalization ability. To address this limitation, we propose a novel generalized lightness adaptation algorithm that extends conventional normalization techniques through a channel filtering design, dubbed Channel Selective Normalization (CSNorm). The proposed CSNorm purposely normalizes the statistics of lightness-relevant channels and keeps other channels unchanged, so as to improve feature generalization and discrimination. To optimize CSNorm, we propose an alternating training strategy that effectively identifies lightness-relevant channels. The model equipped with our CSNorm only needs to be trained on one lightness condition and can be well generalized to unknown lightness conditions. Experimental results on multiple benchmark datasets demonstrate the effectiveness of CSNorm in enhancing the generalization ability for the existing lightness adaptation methods. Code is available at https://github.com/mdyao/CSNorm.

Image aesthetics assessment (IAA) evaluates the generic aesthetic quality of images. Due to the subjectivity of IAA, personalized IAA (PIAA) is essential to offering dedicated image retrieval, editing, and recommendation services to individual users. However, existing PIAA approaches are trained under the centralized machine learning paradigm, which exposes sensitive image and rating data. To enhance PIAA in a privacy-preserving manner, we propose the first-of-its-kind Federated Learning-empowered Personalized Image Aesthetics Assessment (FedPIAA) approach with a simple yet effective model structure to capture image aesthetic patterns and personalized user aesthetic preferences. Extensive experimental comparison against eight baselines using the real-world dataset FLICKER-AES demonstrates that FedPIAA outperforms FedAvg by 1.56% under the small support set and by 4.86% under the large support set in terms of Spearman rank-order correlation coefficient between predicted and ground-truth personalized aesthetics scores, while achieving comparable performance with the best non-FL centralized PIAA approaches.

The performance of existing supervised neuron segmentation methods is highly dependent on the number of accurate annotations, especially when applied to large scale electron microscopy (EM) data. By extracting semantic information from unlabeled data, self-supervised methods can improve the performance of downstream tasks, among which the mask image model (MIM) has been widely used due to its simplicity and effectiveness in recovering original information from masked images. However, due to the high degree of structural locality in EM images, as well as the existence of considerable noise, many voxels contain little discriminative information, making MIM pretraining inefficient on the neuron segmentation task. To overcome this challenge, we propose a decision-based MIM that utilizes reinforcement learning (RL) to automatically search for optimal image masking ratio and masking strategy. Due to the vast exploration space, using single-agent RL for voxel prediction is impractical. Therefore, we treat each input patch as an agent with a shared behavior policy, allowing for multi-agent collaboration. Furthermore, this multi-agent model can capture dependencies between voxels, which is beneficial for the downstream segmentation task. Experiments conducted on representative EM datasets demonstrate that our approach has a significant advantage over alternative self-supervised methods on the task of neuron segmentation. Code is available at https://github.com/ydchen0806/dbMiM.

Neural connectivity reconstruction aims to understand the function of biological reconstruction and promote basic scientific research. The intricate morphology and densely intertwined branches make it an extremely challenging task. Most previous best-performing methods adopt affinity learning or metric learning. Nevertheless, they either neglect to model explicit voxel semantics caused by implicit optimization or are hysteresis to spatial information. Furthermore, the inherent locality of 3D CNNs limits modeling long-range dependencies, leading to sub-optimal results. In this work, we propose a coherent and unified Appearance Prompt Vision Transformer (APViT) to integrate affinity and metric learning to exploit the complementarity by learning long-range spatial dependencies. The proposed APViT enjoys several merits. First, the extension continuity-aware attention module aims at constructing hierarchical attention customized for neuron extensibility and slice continuity to learn instance voxel semantic context from a global perspective and utilize continuity priors to enhance voxel spatial awareness. Second, the appearance prompt modulator is responsible for leveraging voxel-adaptive appearance knowledge conditioned on affinity rich in spatial information to instruct instance voxel semantics, exploiting the potential of affinity learning to complement metric learning. Extensive experimental results on multiple challenging benchmarks demonstrate that our APViT achieves consistent improvements with huge flexibility under the same post-processing strategy.

Image fusion aims to generate a high-quality image from multiple images captured under varying conditions. The key problem of this task is to preserve complementary information while filtering out irrelevant information for the fused result. However, existing methods address this problem by leveraging static convolutional neural networks (CNNs), suffering two inherent limitations during feature extraction,i.e., being unable to handle spatial-variant contents and lacking guidance from multiple inputs. In this paper, we propose a novel mutual-guided dynamic network (MGDN) for image fusion, which allows for effective information utilization across different locations and inputs. Specifically, we design a mutual-guided dynamic filter (MGDF) for adaptive feature extraction, composed of a mutual-guided cross-attention (MGCA) module and a dynamic filter predictor, where the former incorporates additional guidance from different inputs and the latter generates spatial-variant kernels for different locations. In addition, we introduce a parallel feature fusion (PFF) module to effectively fuse local and global information of the extracted features. To further reduce the redundancy among the extracted features while simultaneously preserving their shared structural information, we devise a novel loss function that combines the minimization of normalized mutual information (NMI) with an estimated gradient mask. Experimental results on five benchmark datasets demonstrate that our proposed method outperforms existing methods on four image fusion tasks. The code and model are publicly available at: https://github.com/Guanys-dar/MGDN.

Recent interactive segmentation methods iteratively take source image, user guidance and previously predicted mask as the input without considering the invariant nature of the source image. As a result, the process of extracting features from the source image is repeated in each interaction, resulting in substantial computational redundancy. In this work, we propose the Feature Decoupling-Recycling Network (FDRN), which decouples the modeling components based on their intrinsic discrepancies and then recycles components that can be reused for each user interaction. Thus, the efficiency of the whole interactive process can be significantly improved. To be specific, we apply the Decoupling-Recycling strategy from three perspectives to address three types of discrepancies, respectively. First, our model decouples the learning of source image semantics from the encoding of user guidance to process two types of input domains separately. Second, FDRN decouples high-level and low-level features from stratified semantic representations to enhance feature learning. Third, during the encoding of user guidance, current user guidance is decoupled from historical guidance to highlight the effect of current user guidance. We conduct extensive experiments on 6 datasets from different domains and modalities, which demonstrate the following merits of our model: 1) superior efficiency than other methods, particularly advantageous in the challenging scenarios requiring long-term interactions (up to 4.25x faster), while achieving favorable segmentation performance; 2) strong applicability to various methods serving as a universal enhancement technique; 3) well cross-task generalizability, e.g., to medical image segmentation, and robustness against misleading user guidance.

Normalization techniques that capture image style by statistical representation have become a popular component in deep neural networks. Although image enhancement can be considered as a form of style transformation, there has been little exploration of how normalization affect the enhancement performance. To fully leverage the potential of normalization, we present a novel Transition-Constant Normalization (TCN) for various image enhancement tasks. Specifically, it consists of two streams of normalization operations arranged under an invertible constraint, along with a feature sub-sampling operation that satisfies the normalization constraint. TCN enjoys several merits, including being parameter-free, plug-and-play, and incurring no additional computational costs. We provide various formats to utilize TCN for image enhancement, including seamless integration with enhancement networks, incorporation into encoder-decoder architectures for downsampling, and implementation of efficient architectures. Through extensive experiments on multiple image enhancement tasks, like low-light enhancement, exposure correction, SDR2HDR translation, and image dehazing, our TCN consistently demonstrates performance improvements. Besides, it showcases extensive ability in other tasks including pan-sharpening and medical segmentation. The code is available at \textit{\textcolor{blue}{https://github.com/huangkevinj/TCNorm}}.

Active confocal non-line-of-sight (NLOS) imaging has successfully enabled seeing around corners relying on high-quality transient measurements. However, acquiring spatial-dense transient measurement is time-consuming, raising the question of how to reconstruct satisfactory results from under-scanning measurements (USM). The existing solutions, involving the traditional algorithms, however, are hindered by unsatisfactory results or long computing times. To this end, we propose the first deep-learning-based approach to NLOS imaging from USM. Our proposed end-to-end network is composed of two main components: the transient recovery network (TRN) and the volume reconstruction network (VRN). Specifically, TRN takes the under-scanning measurements as input, utilizes a multiple kernel feature extraction module and a multiple feature fusion module, and outputs sufficient-scanning measurements at the high-spatial resolution. Afterwards, VRN incorporates the linear physics prior of the light-path transport model and reconstructs the hidden volume representation. Besides, we introduce regularized constraints that enhance the perception of more local details while suppressing smoothing effects. The proposed method achieves superior performance on both synthetic data and public real-world data, as demonstrated by extensive experimental results with different under-scanning grids. Moreover, the proposed method delivers impressive robustness at an extremely low scanning grid (i.e., 8$\times$8) and offers high-speed inference (i.e., 50 times faster than the existing iterative solution).

Photon-efficient imaging, which captures 3D images with single-photon sensors, has enabled a wide range of applications. However, two major challenges limit the reconstruction performance, i.e., the low photon counts accompanied by low signal-to-background ratio (SBR) and the multiple returns. In this paper, we propose a unified deep neural network that, for the first time, explicitly addresses these two challenges, and simultaneously recovers depth maps and intensity images from photon-efficient measurements. Starting from a general image formation model, our network is constituted of one encoder, where a non-local block is utilized to exploit the long-range correlations in both spatial and temporal dimensions of the raw measurement, and two decoders, which are designed to recover depth and intensity, respectively. Meanwhile, we investigate the statistics of the background noise photons and propose a noise prior block to further improve the reconstruction performance. The proposed network achieves decent reconstruction fidelity even under extremely low photon counts / SBR and heavy blur caused by the multiple-return effect, which significantly surpasses the existing methods. Moreover, our network trained on simulated data generalizes well to real-world imaging systems, which greatly extends the application scope of photon-efficient imaging in challenging scenarios with a strict limit on optical flux. Code is available at https://github.com/JiayongO-O/PENonLocal.

Self-supervised denoising frameworks have recently been proposed to learn denoising models without noisy-clean image pairs, showing great potential in various applications. The denoising model is expected to produce visually pleasant images without noise patterns. However, it is non-trivial to achieve this goal using self-supervised methods because 1) the self-supervised model is difficult to restore the perceptual information due to the lack of clean supervision, and 2) perceptual quality is relatively subjective to users’ preferences. In this paper, we make the first attempt to build an interactive self-supervised denoising model to tackle the aforementioned problems. Specifically, we propose an interactive two-branch network to effectively restore perceptual information. The network consists of a denoising branch and an interactive branch, where the former focuses on efficient denoising, and the latter modulates the denoising branch. Based on the delicate architecture design, our network can produce various denoising outputs, allowing the user to easily select the most appealing outcome for satisfying the perceptual requirement. Moreover, to optimize the network with only noisy images, we propose a novel two-stage training strategy in a self-supervised way. Once the network is optimized, it can be interactively changed between noise reduction and texture restoration, providing more denoising choices for users. Existing self-supervised denoising methods can be integrated into our method to be user-friendly with interaction. Extensive experiments and comprehensive analyses are conducted to validate the effectiveness of the proposed method.

Light field (LF) cameras suffer from a fundamental trade-off between spatial and angular resolutions. Additionally, due to the significant amount of data that needs to be recorded, the Lytro ILLUM, a modern LF camera, can only capture three frames per second. In this paper, we consider space-time super-resolution (SR) for LF videos, aiming at generating high-resolution and high-frame-rate LF videos from low-resolution and low-frame-rate observations. Extending existing space-time video SR methods to this task directly will meet two key challenges: 1) how to re-organize sub-aperture images (SAIs) efficiently and effectively given highly redundant LF videos, and 2) how to aggregate complementary information between multiple SAIs and frames considering the coherence in LF videos. To address the above challenges, we propose a novel framework for space-time super-resolving LF videos for the first time. First, we propose a novel Multi-Scale Dilated SAI Re-organization strategy for re-organizing SAIs into auxiliary view stacks with decreasing resolution as the Chebyshev distance in the angular dimension increases. In particular, the auxiliary view stack with original resolution preserves essential visual details, while the down-scaled view stacks capture long-range contextual information. Second, we propose the Multi-Scale Aggregated Feature extractor and the Angular-Assisted Feature Interpolation module to utilize and aggregate information from the spatial, angular, and temporal dimensions in LF videos. The former aggregates similar contents from different SAIs and frames for subsequent reconstruction in a disparity-free manner at the feature level, whereas the latter interpolates intermediate frames temporally by implicitly aggregating geometric information. Compared to other potential approaches, experimental results demonstrate that the reconstructed LF videos generated by our framework achieve higher reconstruction quality and better preserve the LF parallax structure and temporal consistency. The implementation code is available at https://github.com/zeyuxiao1997/LFSTVSR.

Unsupervised deformable image registration benefits from progressive network structures such as Pyramid and Cascade. However, existing progressive networks only consider the single-scale deformation field in each level or stage and ignore the long-term connection across non-adjacent levels or stages. In this paper, we present a novel unsupervised learning approach named Self-Distilled Hierarchical Network (SDHNet). By decomposing the registration procedure into several iterations, SDHNet generates hierarchical deformation fields (HDFs) simultaneously in each iteration and connects different iterations utilizing the learned hidden state. Specifically, hierarchical features are extracted to generate HDFs through several parallel gated recurrent units, and HDFs are then fused adaptively conditioned on themselves as well as contextual features from the input image. Furthermore, different from common unsupervised methods that only apply similarity loss and regularization loss, SDHNet introduces a novel self-deformation distillation scheme. This scheme distills the final deformation field as the teacher guidance, which adds constraints for intermediate deformation fields on deformation-value and deformation-gradient spaces respectively. Experiments on five benchmark datasets, including brain MRI and liver CT, demonstrate the superior performance of SDHNet over state-of-the-art methods with a faster inference speed and a smaller GPU memory. Code is available at https://github.com/Blcony/SDHNet.

In this paper, we present the results of the MitoEM challenge on mitochondria 3D instance segmentation from electron microscopy images, organized in conjunction with the IEEE-ISBI 2021 conference. Our benchmark dataset consists of two large-scale 3D volumes, one from human and one from rat cortex tissue, which are 1,986 times larger than previously used datasets. At the time of paper submission, 257 participants had registered for the challenge, 14 teams had submitted their results, and six teams participated in the challenge workshop. Here, we present eight top-performing approaches from the challenge participants, along with our own baseline strategies. Posterior to the challenge, annotation errors in the ground truth were corrected without altering the final ranking. Additionally, we present a retrospective evaluation of the scoring system which revealed that: 1) challenge metric was permissive with the false positive predictions; and 2) size-based grouping of instances did not correctly categorize mitochondria of interest. Thus, we propose a new scoring system that better reflects the correctness of the segmentation results. Although several of the top methods are compared favorably to our own baselines, substantial errors remain unsolved for mitochondria with challenging morphologies. Thus, the challenge remains open for submission and automatic evaluation, with all volumes available for download.

Volumetric image compression has become an urgent task to effectively transmit and store images produced in biological research and clinical practice. At present, the most commonly used volumetric image compression methods are based on wavelet transform, such as JP3D. However, JP3D employs an ideal, separable, global, and fixed wavelet basis to convert input images from pixel domain to frequency domain, which seriously limits its performance. In this paper, we first design a 3-D trained wavelet-like transform to enable signal-dependent and non-separable transform. Then, an affine wavelet basis is introduced to capture the various local correlations in different regions of volumetric images. Furthermore, we embed the proposed wavelet-like transform to an end-to-end compression framework called aiWave to enable an adaptive compression scheme for various datasets. Last but not least, we introduce the weight sharing strategies of the affine wavelet-like transform according to the volumetric data characteristics in the axial direction to reduce the number of parameters. The experimental results show that: 1) when cooperating our trained 3-D affine wavelet-like transform with a simple factorized entropy coding module, aiWave performs better than JP3D and is comparable in terms of encoding and decoding complexities; 2) when adding a context module to remove signal redundancy further, aiWave can achieve a much better performance than HEVC.

Stereo cameras are now commonly used in more and more devices. Nevertheless, visually unpleasant images captured under low-light conditions hinder their practical application. As an initial attempt at low-light stereo image enhancement, we propose a novel Dual-View Enhancement Network (DVENet) based on the Retinex theory, which consists of two stages. The first stage estimates an illumination map to obtain a coarse enhancement result, which boosts the correlation of two views, while the second stage recovers details by integrating the information from two views to achieve fine image quality improvement with the guidance of the illumination map. To fully utilize the dual-view correlation, we further design a wavelet-based view transfer module to efficiently carry out multi-scale detail recovery. Then, we design an illumination-aware attention fusion module to exploit the complementarity between the fused features from two views and the single-view features. Experiments on both synthetic and real-world stereo datasets demonstrate the superiority of our proposed method over existing solutions. The code and model are publicly available at:https://github.com/KevinJ-Huang/Stereo-Low-Light.

Domain generalization aims to generalize a network trained on multiple domains to unknown yet related domains. Operating under the assumption that invariant information generalizes well to unknown domains, previous work has aimed to minimize the discrepancies amongst distributions across given domains. However, without prior regularization of feature distributions, the network in practice overfits the invariant information in the given domains. Moreover, if there are insufficient samples in given domains, then domain generalizability is limited, as diverse domain variations are not captured. To address these two drawbacks, we propose to explicitly map features in known and unknown domains onto latent space in a fixed Gaussian mixture distribution by variational coding. As a result, features in different classes follow Gaussian distributions with different mean values. The predefined latent space narrows discrepancies between known and unknown domains and effectively separates samples into different classes. Moreover, we propose to perturb sample features with gradients from the distribution regularized loss. This perturbation generates samples beyond but near the latent space of prior distributions, which has a profound impact on domain variations. Experiments and visualizations demonstrate the effectiveness of our proposed method.

With the popularization of ultra high definition (UHD) high dynamic range (HDR) displays, recent works focus onupgradinghigh definition (HD) standard dynamic range (SDR) videos to UHD-HDR versions, aiming to provides richer details and higher contrasts on advanced modern displays. However, joint considering theupgrading&downgradingtranslations between two types of videos, which is practical in real applications, is generally neglected. On the one hand,downgradingtranslation is the key to showing UHD-HDR videos on HD-SDR displays. On the other hand, considering both translations enables joint optimization and results in high quality translation. To this end, we propose the bidirectional translation network (BiT-Net), which jointly considers two translations in one network for the first time. In brief, BiT-Net is elaborately designed in aninvertiblefashion that can be efficiently inferred along forward and backward directions fordowngradingandupgradingtasks, respectively. Based on this framework, we divide each direction into three sub-tasks,i.e., decomposition, structure-guided translation, and synthesis, to effectively translate the dynamic range and the high-frequency details. Benefiting from the dedicated architecture, our BiT-Net can work on 1) downgrading UHD-HDR videos, 2) upgrading existing HD-SDR videos, and 3) synthesizing UHD-HDR versions from the downgraded HD-SDR videos. Experiments show that the proposed method achieves state-of-the-art performances on all these three tasks.

Domain generalization aims at generalizing the network trained on multiple domains to unknown but related domains. Under the assumption that different domains share the same classes, previous works can build relationships across domains. However, in realistic scenarios, the change of domains is always followed by the change of categories, which raises a difficulty for collecting sufficient aligned categories across domains. Bearing this in mind, this article introduces union domain generalization (UDG) as a new domain generalization scenario, in which the label space varies across domains, and the categories in unknown domains belong to the union of all given domain categories. The absence of categories in given domains is the main obstacle to aligning different domain distributions and obtaining domain-invariant information. To address this problem, we propose category-stitch learning (CSL), which aims at jointly learning the domain-invariant information and completing missing categories in all domains through an improved variational autoencoder and generators. The domain-invariant information extraction and sample generation cross-promote each other to better generalizability. Additionally, we decouple category and domain information and propose explicitly regularizing the semantic information by the classification loss with transferred samples. Thus our method can breakthrough the category limit and generate samples of missing categories in each domain. Extensive experiments and visualizations are conducted on MNIST, VLCS, PACS, Office-Home, and DomainNet datasets to demonstrate the effectiveness of our proposed method.

The capability of image semantic segmentation may be deteriorated due to the noisy input image, where image denoising prior to segmentation may help. Both image denoising and semantic segmentation have been developed significantly with the advance of deep learning. In this work, we are interested in the synergy between these two tasks by using a holistic deep model. We observe that not only denoising helps combat the drop of segmentation accuracy due to the noisy input, but also pixel-wise semantic information boosts the capability of denoising. We then propose a boosting network to perform denoising and segmentation alternately. The proposed network is composed of multiple segmentation and denoising blocks (SDBs), each of which estimates a semantic map and then uses the map to regularize denoising. Experimental results show that the denoised image quality is improved substantially and the segmentation accuracy is improved to close to that on clean images, and segmentation and denoising are both boosted as the number of SDBs increases. On the Cityscapes dataset, using three SDBs improves the denoising quality to 34.42 dB in PSNR, and the segmentation accuracy to 66.5 in mIoU, when the additive white Gaussian noise level is 50.

Homogeneous instance segmentation aims to identify each instance in an image where all interested instances belong to the same category, such as plant leaves and microscopic cells. Recently, proposal-free methods, which straightforwardly generate instance-aware information to group pixels into different instances, have received increasing attention due to their efficient pipeline. However, they often fail to distinguish adjacent instances due to similar appearances, dense distribution and ambiguous boundaries of instances in homogeneous images. In this paper, we propose a pixel-embedded affinity modeling method for homogeneous instance segmentation, which is able to preserve the semantic information of instances and improve the distinguishability of adjacent instances. Instead of predicting affinity directly, we propose a self-correlation module to explicitly model the pairwise relationships between pixels, by estimating the similarity between embeddings generated from the input image through CNNs. Based on the self-correlation module, we further design a cross-correlation module to maintain the semantic consistency between instances. Specifically, we map the transformed input images with different views and appearances into the same embedding space, and then mutually estimate the pairwise relationships of embeddings generated from the original input and its transformed variants. In addition, to integrate the global instance information, we introduce an embedding pyramid module to model affinity on different scales. Extensive experiments demonstrate the versatile and superior performance of our method on three representative datasets. Code and models are available at https://github.com/weih527/Pixel-Embedded-Affinity.

Deep Convolutional Neural Networks (CNNs) based methods have achieved significant breakthroughs in the task of single image shadow removal. However, the performance of these methods remains limited for several reasons. First, the existing shadow illumination model ignores the spatially variant property of the shadow images, hindering their further performance. Second, most deep CNNs based methods directly estimate the shadow free results from the input shadow images like a black box, thus losing the desired interpretability. To address these issues, we first propose a new shadow illumination model for the shadow removal task. This new shadow illumination model ensures the identity mapping among unshaded regions, and adaptively performs fine grained spatial mapping between shadow regions and their references. Then, based on the shadow illumination model, we reformulate the shadow removal task as a variational optimization problem. To effectively solve the variational problem, we design an iterative algorithm and unfold it into a deep network, naturally increasing the interpretability of the deep model. Experiments show that our method could achieve SOTA performance with less than half parameters, one-fifth of floating-point of operations (FLOPs), and over seventeen times faster than SOTA method (DHAN).

In this paper, we study the problem of stereo matching from a pair of images with different resolutions, e.g., those acquired with a tele-wide camera system. Due to the difficulty of obtaining ground-truth disparity labels in diverse real-world systems, we start from an unsupervised learning perspective. However, resolution asymmetry caused by unknown degradations between two views hinders the effectiveness of the generally assumed photometric consistency. To overcome this challenge, we propose to impose the consistency between two views in a feature space instead of the image space, named feature-metric consistency. Interestingly, we find that, although a stereo matching network trained with the photometric loss is not optimal, its feature extractor can produce degradation-agnostic and matching-specific features. These features can then be utilized to formulate a feature-metric loss to avoid the photometric inconsistency. Moreover, we introduce a self-boosting strategy to optimize the feature extractor progressively, which further strengthens the feature-metric consistency. Experiments on both simulated datasets with various degradations and a self-collected real-world dataset validate the superior performance of the proposed method over existing solutions.

Previous LiDAR scene flow estimation methods, especially recurrent neural networks, usually suffer from structure distortion in challenging cases, such as sparse reflection and motion occlusions. In this paper, we propose a novel optimization method based on a recurrent neural network to predict LiDAR scene flow in a weakly supervised manner. Specifically, our neural recurrent network exploits direct rigidity constraints to preserve the geometric structure of the warped source scene during an iterative alignment procedure. An error awarded optimization strategy is proposed to update the LiDAR scene flow by minimizing the point measurement error instead of reconstructing the cost volume multiple times. Trained on two autonomous driving datasets, our network outperforms recent state-of-the-art networks on lidarKITTI by a large margin. The code and models will be available at https://github.com/gtdong-ustc/LiDARSceneFlow.

Images captured with improper exposures usually bring unsatisfactory visual effects. Previous works mainly focus on either underexposure or overexposure correction, resulting in poor generalization to various exposures. An alternative solution is to mix the multiple exposure data for training a single network. However, the procedures of correcting underexposure and overexposure to normal exposures are much different from each other, leading to large discrepancies for the network in correcting multiple-exposures, thus resulting in poor performance. The key point to address this issue lies in bridging different exposure representations. To achieve this goal, we design a multiple exposure correction framework based on an Exposure Normalization and Compensation (ENC) module. Specifically, the ENC module consists of an exposure normalization part for mapping different exposure features to the exposure-invariant feature space, and a compensation part for integrating the initial features unprocessed by the exposure normalization part to ensure the completeness of information. Besides, to further alleviate the imbalanced performance caused by variations in the optimization process, we introduce a parameter regularization fine-tuning strategy to improve the performance of the worst-performed exposure without degrading other exposures. Our model empowered by ENC outperforms the existing methods by more than 2dB and is robust to multiple image enhancement tasks, demonstrating its effectiveness and generalization capability for real-world applications. Code: https://github.com/KevinJ-Huang/ExposureNorm-Compensation.

We study the problem of contextual outpainting, which aims to hallucinate the missing background contents based on the remaining foreground contents. Existing image outpainting methods focus on completing object shapes or extending existing scenery textures, neglecting the semantically meaningful relationship between the missing and remaining contents. To explore the semantic cues provided by the remaining foreground contents, we propose a novel ConTextual Outpainting GAN (CTO-GAN), leveraging the semantic layout as a bridge to synthesize coherent and diverse background contents. To model the contextual correlation between foreground and background contents, we incorporate an object-level contrastive loss to regularize the learning of cross-modal representations of foreground contents and the corresponding background semantic layout, facilitating accurate semantic reasoning. Furthermore, we improve the realism of the generated background contents via detecting generated context in adversarial training. Extensive experiments demonstrate that the proposed method achieves superior performance compared with existing solutions on the challenging COCO-stuff dataset. Project page: https://ddlee-cn.github.io/cto-gan.

Space-time memory (STM) based video object segmentation (VOS) networks usually keep increasing memory bank every several frames, which shows excellent performance. However, 1) the hardware cannot withstand the ever-increasing memory requirements as the video length increases. 2) Storing lots of information inevitably introduces lots of noise, which is not conducive to reading the most important information from the memory bank. In this paper, we propose a Recurrent Dynamic Embedding (RDE) to build a memory bank of constant size. Specifically, we explicitly generate and update RDE by the proposed Spatio-temporal Aggregation Module (SAM), which exploits the cue of historical information. To avoid error accumulation owing to the recurrent usage of SAM, we propose an unbiased guidance loss during the training stage, which makes SAM more robust in long videos. Moreover, the predicted masks in the memory bank are inaccurate due to the inaccurate network inference, which affects the seg-mentation of the query frame. To address this problem, we design a novel self-correction strategy so that the network can repair the embeddings of masks with different qualities in the memory bank. Extensive experiments show our method achieves the best tradeoff between performance and speed. Code is available at https://github.com/Limingxing00/RDE-VOS-CVPR2022.

Diffractive snapshot hyperspectral imaging based on the deep optics framework has been striving to capture the spectral images of dynamic scenes. However, existing deep optics frameworks all suffer from the mismatch between the optical hardware and the reconstruction algorithm due to the quantization operation in the diffractive optical element (DOE) fabrication, leading to the limited performance of hyperspectral imaging in practice. In this paper, we propose the quantization-aware deep optics for diffractive snapshot hyperspectral imaging. Our key observation is that common lithography techniques used in fabricating DOEs need to quantize the DOE height map to a few levels, and can freely set the height for each level. Therefore, we propose to integrate the quantization operation into the DOE height map optimization and design an adaptive mechanism to adjust the physical height of each quantization level. According to the optimization, we fabricate the quantized DOE directly and build a diffractive hyperspectral snapshot imaging system. Our method develops the deep optics framework to be more practical through the awareness of and adaptation to the quantization operation of the DOE physical structure, making the fabricated DOE and the reconstruction algorithm match each other systematically. Extensive synthetic simulation and real hardware experiments validate the superior performance of our method.

Recent advanced proposal-free instance segmentation methods have made significant progress in biological images. However, existing methods are vulnerable to local imaging artifacts and similar object appearances, resulting in over-merge and over-segmentation. To reduce these two kinds of errors, we propose a new biological instance segmentation framework based on a superpixel-guided graph, which consists of two stages, i.e., superpixel-guided graph construction and superpixel agglomeration. Specifically, the first stage generates enough superpixels as graph nodes to avoid over-merge, and extracts node and edge features to construct an initialized graph. The second stage agglomerates superpixels into instances based on the relationship of graph nodes predicted by a graph neural network (GNN). To solve over-segmentation and prevent introducing additional over-merge, we specially design two loss functions to supervise the GNN, i.e., a repulsion-attraction (RA) loss to better distinguish the relationship of nodes in the feature space, and a maximin agglomeration score (MAS) loss to pay more attention to crucial edge classification. Extensive experiments on three representative biological datasets demonstrate the superiority of our method over existing state-of-the-art methods. Code is available at https://github.com/liuxy1103/BISSG.

This paper proposes a new "decompose-and-edit" paradigm for the text-based speech insertion task that facilitates arbitrarylength speech insertion and even full sentence generation.In the proposed paradigm, global and local factors in speech are explicitly decomposed and separately manipulated to achieve high speaker similarity and continuous prosody.Specifically, we proposed to represent the global factors by multiple tokens, which are extracted by cross-attention operation and then injected back by link-attention operation.Due to the rich representation of global factors, we manage to achieve high speaker similarity in a zero-shot manner.In addition, we introduce a prosody smoothing task to make the local prosody factor context-aware and therefore achieve satisfactory prosody continuity.We further achieve high voice quality with an adversarial training stage.In the subjective test, our method achieves state-of-the-art performance in both naturalness and similarity.Audio samples can be found at https://ydcustc.github.io/retrieverTTS-demo/.

Images captured under improper exposures including underexposure and overexposure often suffer from unsatisfactory visual effects. Since their correction procedures are quite different, it is challenging for a single network to correct various exposures. The key to addressing this issue is consistently learning underexposure and overexposure corrections. To achieve this goal, we propose an Exposure-Consistency Processing (ECP) module to consistently learn the representation of both underexposure and overexposure in the feature space. Specifically, the ECP module employs the bilateral activation mechanism that derives both underexposure and overexposure property features for exposure-consistency representation modeling, which is followed by two shared-weight branches to process these features. Based on the ECP module, we build the whole network by utilizing it as the basic unit. Additionally, to further assist the exposure-consistency learning, we develop an Exposure-Consistency Constraining (ECC) strategy that augments the various local region exposures and then constrains the feature representation change between the exposure augmented image and the original one. Our proposed network is lightweight and outperforms existing methods remarkably, while the ECP module can also be extended to other baselines, demonstrating its superiority and scalability. code: https://github.com/KevinJ-Huang/ECLNet.

Estimating scene flow for point clouds is one of the key problems in 3D scene understanding and autonomous driving. Recently the point transformer architecture has become a popular and successful solution for 3D computer vision tasks, e.g., point cloud object detection and completion, but its application to scene flow estimation is rarely explored. In this work, we provide a full transformer based solution for scene flow estimation. We first introduce a novel relative position guided point attention mechanism. Then to relax the memory consumption in practice, we provide an efficient implementation of our proposed point attention layer via matrix factorization and nearest neighbor sampling. Finally, we build a pyramid transformer, named RPPformer-Flow, to estimate the scene flow between two consecutive point clouds in a coarse-to-fine manner. We evaluate our RPPformer-Flow on the FlyingThings3D and KITTI Scene Flow 2015 benchmarks. Experimental results show that our method outperforms previous state-of-the-art methods with large margins.

Magnetic resonance imaging (MRI) with high resolution (HR) provides more detailed information for accurate diagnosis and quantitative image analysis. Despite the significant advances, most existing super-resolution (SR) reconstruction network for medical images has two flaws: 1) All of them are designed in a black-box principle, thus lacking sufficient interpretability and further limiting their practical applications. Interpretable neural network models are of significant interest since they enhance the trustworthiness required in clinical practice when dealing with medical images. 2) most existing SR reconstruction approaches only use a single contrast or use a simple multi-contrast fusion mechanism, neglecting the complex relationships between different contrasts that are critical for SR improvement. To deal with these issues, in this paper, a novel Model-Guided interpretable Deep Unfolding Network (MGDUN) for medical image SR reconstruction is proposed. The Model-Guided image SR reconstruction approach solves manually designed objective functions to reconstruct HR MRI. We show how to unfold an iterative MGDUN algorithm into a novel model-guided deep unfolding network by taking the MRI observation matrix and explicit multi-contrast relationship matrix into account during the end-to-end optimization. Extensive experiments on the multi-contrast IXI dataset and BraTs 2019 dataset demonstrate the superiority of our proposed model.

Images captured in clinic such as MRI scans are usually in 3D formats with high spatial resolutions. Existing learning-based models for medical image registration consume large GPU memories and long inference time, which is difficult to be deployed in resource-limited diagnosis scenarios. To address this problem, instead of shrinking the model size as in previous works, we turn to reducing the input resolution of existing registration models and boosting their performance through knowledge distillation. Specifically, we propose a cross-resolution distillation (CRD) scheme, which is designed to train low-resolution models under the guidance of corresponding high-resolution models. Nevertheless, due to the resolution gap between features in high/low-resolution models, straightforward distillation is difficult to apply. To overcome this challenge, we first introduce a feature-shifted teacher (FST) to shift and fuse features of high/low-resolution models. Then, we exploit this teacher model to guide the learning of the low-resolution student model with distillation losses on both features and deformation fields. Finally, we only need to use the distilled student model during inference. Experimental results on four 3D medical image datasets demonstrate that the low-resolution models trained through our CRD scheme use fewer than 20% GPU memories and less than 20% inference time while achieving competitive performance compared with corresponding high-resolution models.

Despite impressive progress made by recent image inpainting methods, they often fail to predict the original content when the corrupted region contains unique structures, especially for landmark images. Applying similar images as a reference is helpful but introduces a style gap of textures, resulting in color misalignment. To this end, we propose a style-robust approach for reference-guided landmark image inpainting, taking advantage of both the representation power of learned deep features and the structural prior from the reference image. By matching deep features, our approach builds style-robust nearest-neighbor mapping vector fields between the corrupted and reference images, in which the loss of information due to corruption leads to mismatched mapping vectors. To correct these mismatched mapping vectors based on the relationship between the uncorrupted and corrupted regions, we introduce mutual nearest neighbors as reliable anchors and interpolate around these anchors progressively. Finally, based on the corrected mapping vector fields, we propose a two-step warping strategy to complete the corrupted image, utilizing the reference image as a structural “blueprint”, avoiding the style misalignment problem. Extensive experiments show that our approach effectively and robustly assists image inpainting methods in restoring unique structures in the corrupted image.

Geometry-based point cloud compression (G-PCC) can achieve remarkable compression efficiency for point clouds. However, it still leads to serious attribute compression artifacts, especially under low bitrate scenarios. In this paper, we propose a Multi-Scale Graph Attention Network (MS-GAT) to remove the artifacts of point cloud attributes compressed by G-PCC. We first construct a graph based on point cloud geometry coordinates and then use the Chebyshev graph convolutions to extract features of point cloud attributes. Considering that one point may be correlated with points both near and far away from it, we propose a multi-scale scheme to capture the short- and long-range correlations between the current point and its neighboring and distant points. To address the problem that various points may have different degrees of artifacts caused by adaptive quantization, we introduce the quantization step per point as an extra input to the proposed network. We also incorporate a weighted graph attentional layer into the network to pay special attention to the points with more attribute artifacts. To the best of our knowledge, this is the first attribute artifacts removal method for G-PCC. We validate the effectiveness of our method over various point clouds. Objective comparison results show that our proposed method achieves an average of 9.74% BD-rate reduction compared with Predlift and 10.13% BD-rate reduction compared with RAHT. Subjective comparison results present that visual artifacts such as color shifting, blurring, and quantization noise are reduced.

Deformation decomposition serves as a good solution for deformable image registration when the deformation is large. Current deformation decomposition methods can be categorized into cascade-based methods and pyramid-based methods. However, cascade-based methods suffer from heavy computational burdens and long inference time due to their structures of repeated subnetworks, while the effectiveness of pyramid-based methods is constrained by their limited numbers of resolution levels. In this paper, to address both the insufficient and inefficient decomposition problems in current deformation decomposition methods, we propose a recursive decomposition network (RDN) to offer a novel solution for deformable image registration. Stage-wise recursion can efficiently decompose a large deformation into different pyramid estimation stages without using repeated subnetworks like in cascade-based methods. Level-wise recursion can sufficiently decompose the deformation inside each resolution level instead of only one-time estimation like in pyramid-based methods. Extensive experiments and ablation studies on two representative datasets validate the effectiveness and efficiency of our proposed RDN.

Serial section transmission electron micro-scopy (ssTEM) reveals biological information at a scale of nanometer and plays an important role in the ultrastructural analysis. However, due to the imperfect preparation of biological samples, ssTEM images are usually degraded with various artifacts that greatly challenge the subsequent analysis and visualization. In this paper, we introduce a unified deep learning framework for ssTEM image restoration which addresses three main types of artifacts, i.e., Support Film Folds (SFF), Staining Precipitates (SP), and Missing Sections (MS). To achieve this goal, we first model the appearance of SFF and SP artifacts by conducting comprehensive analyses on the statistics of real degraded images, relying on which we can then simulate a large number of paired images (degraded/artifacts-free) for training a deep restoration network. Then, we design a coarse-to-fine restoration network consisting of three modules, i.e., interpolation, correction, and fusion. The interpolation module exploits the adjacent artifacts-free images for an initial restoration, while the correction module resorts to the degraded image itself to rectify the artifacts. Finally, the fusion module jointly utilizes the above two results to further improve the restoration fidelity. Experimental results on both synthetic and real test data validate the significantly improved performance of our proposed framework over existing solutions, in terms of both image restoration fidelity and neuron segmentation accuracy. To the best of our knowledge, this is the first unified deep learning framework for ssTEM image restoration from different types of artifacts. Code is available at https://github.com/sydeng99/ssTEM-restoration.

Emerging deep learning-based methods have enabled great progress in automatic neuron segmentation from Electron Microscopy (EM) volumes. However, the success of existing methods is heavily reliant upon a large number of annotations that are often expensive and time-consuming to collect due to dense distributions and complex structures of neurons. If the required quantity of manual annotations for learning cannot be reached, these methods turn out to be fragile. To address this issue, in this article, we propose a two-stage, semi-supervised learning method for neuron segmentation to fully extract useful information from unlabeled data. First, we devise a proxy task to enable network pre-training by reconstructing original volumes from their perturbed counterparts. This pre-training strategy implicitly extracts meaningful information on neuron structures from unlabeled data to facilitate the next stage of learning. Second, we regularize the supervised learning process with the pixel-level prediction consistencies between unlabeled samples and their perturbed counterparts. This improves the generalizability of the learned model to adapt diverse data distributions in EM volumes, especially when the number of labels is limited. Extensive experiments on representative EM datasets demonstrate the superior performance of our reinforced consistency learning compared to supervised learning, i.e., up to 400% gain on the VOI metric with only a few available labels. This is on par with a model trained on ten times the amount of labeled data in a supervised manner. Code is available at https://github.com/weih527/SSNS-Net.

The large data volume of point clouds poses severe challenges for efficient storage and transmission in recent years. In this paper, we propose the first--to our best knowledge--end-to-end deep framework for compressing point cloud attributes. Specifically, we propose a point cloud lossy attribute autoencoder, which directly encodes and decodes point cloud attributes with the help of geometry, instead of voxelizing or projecting the points. In the autoencoder, we propose a second-order point convolution that utilizes the spatial correlations between more points and the nonlinear relationship between attribute features. We introduce a dense point-inception block, which derives from a combination of an inception-style block and a dense block, to improve feature propagation. In addition, we devise a multiscale loss to guide the autoencoder in focusing attention on the coarse-grained points with better coverage of the entire point cloud, which makes it easier for the autoencoder to obtain better optimization of the qualities of all points. Experimental results show that our proposed framework still has a performance gap compared with the state-of-the-art algorithms in the MPEG G-PCC reference software TMC13. However, it does outperform the RAHT-RLGR, which is one of the core transforms used in TMC13 without many well-designed techniques that make TMC13 what it is today. It outperforms RAHT-RLGR by 2.63 dB, 1.77 dB, and 3.40 dB on average in terms of the BD-PSNR for the Y, U, and V components. A subjective quality comparison demonstrates that our framework can preserve more textures and reduce blocking and color noise artifacts.

The fast development of neuromorphic hardwares promotes Spiking Neural Networks (SNNs) to a thrilling research avenue. Current SNNs, though much efficient, are less effective compared with leading Artificial Neural Networks (ANNs) especially in supervised learning tasks. Recent efforts further demonstrate the potential of SNNs in supervised learning by introducing approximated backpropagation (BP) methods. To deal with the non-differentiable spike function in SNNs, these BP methods utilize information from the spatio-temporal domain to adjust the model parameters. With the increasing of time window and network size, the computational complexity of spatio-temporal backpropagation augments dramatically. In this paper, we propose a new backpropagation method for SNNs based on the accumulated spiking flow (ASF), i.e. ASF-BP. In the proposed ASF-BP method, updating parameters does not rely on the spike train of spiking neurons but leverage accumulated inputs and outputs of spiking neurons over the time window, which reduces the BP complexity significantly. We further present an adaptive linear estimation model to approach the dynamic characteristics of spiking neurons statistically. Experimental results demonstrate that with our proposed ASF-BP method, light-weight convolutional SNNs achieve superior performances compared with other spike-based BP methods on both non-neuromorphic (MNIST, CIFAR10) and neuromorphic (CIFAR10-DVS) datasets. The code is available at https://github.com/neural-lab/ASF-BP.

Deep learning provides a new avenue for light field super-resolution (SR). However, the domain gap caused by drastically different light field acquisition conditions poses a main obstacle in practice. To fill this gap, we propose a zero-shot learning framework for light field SR, which learns a mapping to super-resolve the reference view with examples extracted solely from the input low-resolution light field itself. Given highly limited training data under the zero-shot setting, however, we observe that it is difficult to train an end-to-end network successfully. Instead, we divide this challenging task into three sub-tasks, i.e., pre-upsampling, view alignment, and multi-view aggregation, and then conquer them separately with simple yet efficient CNNs. Moreover, the proposed framework can be readily extended to finetune the pre-trained model on a source dataset to better adapt to the target input, which further boosts the performance of light field SR in the wild. Experimental results validate that our method not only outperforms classic non-learning-based methods, but also generalizes better to unseen light fields than state-of-the-art deep-learning-based methods when the domain gap is large.

Inspired by the fact that human eyes continue to develop tracking ability in early and middle childhood, we propose to use tracking as a proxy task for a computer vision system to learn the visual representations. Modelled on the Catch game played by the children, we design a Catch-the-Patch (CtP) game for a 3D-CNN model to learn visual representations that would help with video-related tasks. In the proposed pretraining framework, we cut an image patch from a given video and let it scale and move according to a pre-set trajectory. The proxy task is to estimate the position and size of the image patch in a sequence of video frames, given only the target bounding box in the first frame. We discover that using multiple image patches simultaneously brings clear benefits. We further increase the difficulty of the game by randomly making patches invisible. Extensive experiments on mainstream benchmarks demonstrate the superior performance of CtP against other video pretraining methods. In addition, CtP-pretrained features are less sensitive to domain gaps than those trained by a supervised action recognition task. When both trained on Kinetics-400, we are pleasantly surprised to find that CtP-pretrained representation achieves much higher action classification accuracy than its fully supervised counterpart on Something-Something dataset.

Compact video super-resolution (VSR) networks can be easily deployed on resource-limited devices, e.g., smartphones and wearable devices, but have considerable performance gaps compared with complicated VSR networks that require a large amount of computing resources. In this paper, we aim to improve the performance of compact VSR networks without changing their original architectures, through a knowledge distillation approach that transfers knowledge from a complicated VSR network to a compact one. Specifically, we propose a space-time distillation (STD) scheme to exploit both spatial and temporal knowledge in the VSR task. For space distillation, we extract spatial attention maps that hint the high-frequency video content from both networks, which are further used for transferring spatial modeling capabilities. For time distillation, we narrow the performance gap between compact models and complicated models by distilling the feature similarity of the temporal memory cells, which are encoded from the sequence of feature maps generated in the training clips using ConvLSTM. During the training process, STD can be easily incorporated into any network without changing the original network architecture. Experimental results on standard benchmarks demonstrate that, in resource-constrained situations, the proposed method notably improves the performance of existing VSR networks without increasing the inference time.

Existing speech enhancement methods mainly separate speech from noises at the signal level or in the time-frequency domain. They seldom pay attention to the semantic information of a corrupted signal. In this paper, we aim to bridge this gap by extracting phoneme identities to help speech enhancement. Specifically, we propose a phoneme-based distribution regularization (PbDr) for speech enhancement, which incorporates frame-wise phoneme information into speech enhancement network in a conditional manner. As different phonemes always lead to different feature distributions in frequency, we propose to learn a parameter pair, i.e. scale and bias, through a phoneme classification vector to modulate the speech enhancement network. The modulation parameter pair includes not only frame-wise but also frequency-wise conditions, which effectively map features to phoneme-related distributions. In this way, we explicitly regularize speech enhancement features by recognition vectors. Experiments on public datasets demonstrate that the proposed PbDr module can not only boost the perceptual quality for speech enhancement but also the recognition accuracy of an ASR system on the enhanced speech. This PbDr module could be readily incorporated into other speech enhancement networks as well.

Event cameras, which output events by detecting spatio- temporal brightness changes, bring a novel paradigm to image sensors with high dynamic range and low latency. Previous works have achieved impressive performances on event-based video reconstruction by introducing convolutional neural networks (CNNs). However, intrinsic locality of convolutional operations is not capable of modeling long-range dependency, which is crucial to many vision tasks. In this paper, we present a hybrid CNN- Transformer network for event-based video reconstruction (ET-Net), which merits the fine local information from CNN and global contexts from Transformer In addition, we further propose a Token Pyramid Aggregation strategy to implement multi-scale token integration for relating internal and intersected semantic concepts in the token-space. Experimental results demonstrate that our proposed method achieves superior performance over state-of-the-art methods on multiple real-world event datasets. The code is available at https://github.com/WarranWeng/ET-Net.

Dual-camera systems containing a color camera and a monochrome camera are widely equipped on smartphones. The color camera captures chrominance information while the monochrome camera captures fine details, which causes asymmetry across spectral and spatial dimensions. In these imaging systems, the chrominance information of low-resolution (LR) color images and the spatial information of high-resolution (HR) monochrome images are highly complementary. In this paper, we propose an elaborate convolutional neural network to recover HR color images by transferring color information from LR color images to HR monochrome images. The network contains a novel feature extraction module named U-ASPP and an asymmetric parallax attention module (APAM). Our network achieves state-of-the-art performance on the Flickr1024 stereo dataset with high efficiency. Moreover, the effectiveness of our trained network is validated in real-world asymmetric image pairs captured by a smartphone, which demonstrates that our method has high generalization capability in real-world imaging systems.

Stereo Video Super-Resolution (StereoVSR) aims to generate high-resolution video steams from two low-resolution videos under stereo settings. Existing video super-resolution and stereo image super-resolution techniques can be extended to tackle the StereoVSR task, yet they cannot make full use of the multi-view and temporal information to achieve satisfactory performance. In this paper, we propose a novel Stereo Video Super-Resolution Network (SVSRNet) to fulfill the StereoVSR task via exploiting view-temporal correlations. First, we devise a view-temporal attention module (VTAM) to integrate the information of cross-time-cross-view for constructing high-resolution stereo videos. Second, we propose a spatial-temporal fusion module (STFM), which aggregates the information across time in intra-view to emphasize important features for subsequent restoration. In addition, we design a view-temporal consistency loss function to enforce consistency constraint of superresolved stereo videos. Comprehensive experimental results demonstrate that our method generates superior results.

Albeit existing deep learning-based image de-raining methods have achieved promising results, most of them only extract single scale features, and neglect the fact that similar rain streaks appear repeatedly across different scales. Therefore, this paper aims to explore the cross-scale cues in a multi-scale fashion. Specifically, we first introduce an adaptive-kernel pyramid to provide effective multi-scale information. Then, we design two cross-scale similarity attention blocks (CSSABs) to search spatial and channel relationships between two scales, respectively. The spatial CSSAB explores the spatial similarity between pixels of cross-scale features, while the channel CSSAB emphasizes the interdependencies among cross-scale features. To further improve the diversity of features, we adopt the wavelet transformation and multi-head mechanism in CSSABs to generate multifocal features which focus on different areas. Finally, based on our CSSABs, we construct an effective multifocal attention-based cross-scale network, which exhaustively utilizes the cross-scale correlations of both rain streaks and background, to achieve image de-raining. Experiments show the superiority of our network over state-of-the-art image de-raining approaches both qualitatively and quantitatively. The source code and pre-trained models are available at https://github.com/zhangzheyu0/Multifocal_derain.

Generative adversarial networks (GANs) have been extensively used for training networks that perform image generation. After training, the discriminator in GAN was not used anymore. We propose to recycle the trained discriminator for another use: no-reference image quality assessment (NR-IQA). We are motivated by twofold facts. First, in Wasserstein GAN (WGAN), the discriminator is designed to calculate the distance between the distribution of generated images and that of real images; thus, the trained discriminator may encode the distribution of real-world images. Second, NR-IQA often needs to leverage the distribution of real-world images for assessing image quality. We then conjecture that using the trained discriminator for NR-IQA may help get rid of any human-labeled quality opinion scores and lead to a new opinion-unaware (OU) method. To validate our conjecture, we start from a restricted NR-IQA problem, that is IQA for artificially super-resolved images. We train super-resolution (SR) WGAN with two kinds of discriminators: one is to directly evaluate the entire image, and the other is to work on small patches. For the latter kind, we obtain patch-wise quality scores, and then have the flexibility to fuse the scores, e.g., by weighted average. Moreover, we directly extend the trained discriminators for authentically distorted images that have different kinds of distortions. Our experimental results demonstrate that the proposed method is comparable to the state-of-the-art OU NR-IQA methods on SR images and is even better than them on authentically distorted images. Our method provides a better interpretable approach to NR-IQA. Our code and models are available at https://github.com/YunanZhu/RecycleD.

Deep learning provides a new avenue for image restoration, which demands a delicate balance between fine-grained details and high-level contextualized information during recovering the latent clear image. In practice, however, existing methods empirically construct encapsulated end-to-end mapping networks without deepening into the rationality, and neglect the intrinsic prior knowledge of restoration task. To solve the above problems, inspired by Taylor’s Approximations, we unfold Taylor’s Formula to construct a novel framework for image restoration. We find the main part and the derivative part of Taylor’s Approximations take the same effect as the two competing goals of high-level contextualized information and spatial details of image restoration respectively. Specifically, our framework consists of two steps, which are correspondingly responsible for the mapping and derivative functions. The former first learns the high-level contextualized information and the later combines it with the degraded input to progressively recover local high-order spatial details. Our proposed framework is orthogonal to existing methods and thus can be easily integrated with them for further improvement, and extensive experiments demonstrate the effectiveness and scalability of our proposed framework.

Deep neural networks (DNNs) have been widely used for stereo depth estimation, which achieve great success in performance. In this paper, we introduce a novel flipping strategy for DNN on the stereo depth estimation task. Specifically, based on a common DNN for stereo matching, we apply the flipping operation for both input stereo images, which are further fed to the original DNN. A flipping loss function is proposed to jointly train the network with the initial loss. We apply our strategy to many representative networks in both supervised and self-supervised manners. Extensive experimental results demonstrate that our proposed strategy improves the performance of these networks.

With the rapid development of whole brain imaging technology, a large number of brain images have been produced, which puts forward a great demand for efficient brain image compression methods. At present, the most commonly used compression methods are all based on 3-D wavelet transform, such as JP3D. However, traditional 3-D wavelet transforms are designed manually with certain assumptions on the signal, but brain images are not as ideal as assumed. What's more, they are not directly optimized for compression task. In order to solve these problems, we propose a trainable 3-D wavelet transform based on the lifting scheme, in which the predict and update steps are replaced by 3-D convolutional neural networks. Then the proposed transform is embedded into an end-to-end compression scheme called iWave3D, which is trained with a large amount of brain images to directly minimize the rate-distortion loss. Experimental results demonstrate that our method outperforms JP3D significantly by 2.012 dB in terms of average BD-PSNR.

Deep learning-based methods especially using convolutional neural network (CNN) and generative adversarial network (GAN) have achieved certain success for the task of image inpainting. The previous methods usually try to generate the content in the missing areas from scratch. However, these methods have difficulty in producing salient image structures that appear natural and consistent with the neighborhood, especially when the missing area is large. In this paper, we address the challenge by introducing edges into the convolutional GAN-based inpainting. We split the inpainting task into two steps: first edge generation, then edge-based image generation. We adopt CNN to accomplish the two steps and use GAN-based training, thus our method is named E2I: generative inpainting from edge to image. Specifically, we adopt a deep network-based edge detector to achieve an edgeness map of an incomplete image, then we fill-in the missing areas in the edgeness map, and finally generate the missing pixels with the assistance of the complete edgeness map. We verify the proposed method on three challenging image datasets: Places2, ImageNet, and CelebA. We also compare our method with the state-of-the-arts on the Places2 test set. Our experimental results demonstrate the superior performance of our method in producing more plausible inpainting results.

Manually labeling neurons from high-resolution but noisy and low-contrast optical microscopy (OM) images is tedious. As a result, the lack of annotated data poses a key challenge when applying deep learning techniques for reconstructing neurons from noisy and low-contrast OM images. While traditional tracing methods provide a possible way to efficiently generate labels for supervised network training, the generated pseudo-labels contain many noisy and incorrect labels, which lead to severe performance degradation. On the other hand, the publicly available dataset, BigNeuron, provides a large number of single 3D neurons that are reconstructed using various imaging paradigms and tracing methods. Though the raw OM images are not fully available for these neurons, they convey essential morphological priors for complex 3D neuron structures. In this paper, we propose a new approach to exploit morphological priors from neurons that have been reconstructed for training a deep neural network to extract neuron signals from OM images. We integrate a deep segmentation network in a generative adversarial network (GAN), expecting the segmentation network to be weakly supervised by pseudo-labels at the pixel level while utilizing the supervision of previously reconstructed neurons at the morphology level. In our morphological-prior-guided neuron reconstruction GAN, named MP-NRGAN, the segmentation network extracts neuron signals from raw images, and the discriminator network encourages the extracted neurons to follow the morphology distribution of reconstructed neurons. Comprehensive experiments on the public VISoR-40 dataset and BigNeuron dataset demonstrate that our proposed MP-NRGAN outperforms state-of-the-art approaches with less training effort.

Deep learning is known to be effective at automating the generation of representations, which eliminates the need for handcrafted features. For the task of personalized recommendation, deep learning-based methods have achieved great success by learning efficient representations of multimedia items, especially images and videos. Previous works usually adopt simple, single-modality representations of user interest, such as user embeddings, which cannot fully characterize the diversity and volatility of user interest. To address this problem, in this paper we focus on learning and fusing multiple kinds of user interest representations by leveraging deep networks. Specifically, we consider efficient representations of four aspects of user interest: first, we use latent representation, i.e. user embedding, to profile the overall interest; second, we propose item-level representation, which is learned from and integrates the features of a user's historical items; third, we investigate neighbor-assisted representation, i.e. using neighboring users' information to characterize user interest collaboratively; fourth, we propose category-level representation, which is learned from the categorical attributes of a user's historical items. In order to integrate these multiple user interest representations, we study both early fusion and late fusion; where for early fusion, we study different fusion functions. We validate the proposed method on two real-world video recommendation datasets for micro-video and movie recommendations, respectively. Experimental results demonstrate that our method outperforms existing state-of-the-arts by a significant margin. Our code is publicly available.

Single-photon avalanche diode (SPAD) has been widely used in active 3D imaging due to its extremely high photon sensitivity and picosecond time resolution. However, long-range active 3D imaging is still a great challenge, since only a few signal photons mixed with strong background noise can return from multiple reflectors of the scene due to the divergence of the light beam and the receiver's field of view (FoV), which would bring considerable distortion and blur to the recovered depth map. In this paper, we propose a deep learning based depth reconstruction method for long range single-photon 3D imaging where the “multiple-returns” issue exists. Specifically, we model this problem as a deblurring task and design a multi-scale convolutional neural network combined with elaborate loss functions, which promote the reconstruction of an accurate depth map with fine details and clear boundaries of objects. The proposed method achieves superior performance over several different sizes of receiver's FoV on a synthetic dataset compared with existing state-of-the-art methods and the trained model under a specific FoV has a strong generalization capability across different sizes of FoV, which is essential for practical applications. Moreover, we conduct outdoor experiments and demonstrate the effectiveness of our method in a real-world long range imaging system.

Time-frequency (T-F) domain masking is a mainstream approach for single-channel speech enhancement. Recently, focuses have been put to phase prediction in addition to amplitude prediction. In this paper, we propose a phase-and-harmonics-aware deep neural network (DNN), named PHASEN, for this task. Unlike previous methods which directly use a complex ideal ratio mask to supervise the DNN learning, we design a two-stream network, where amplitude stream and phase stream are dedicated to amplitude and phase prediction. We discover that the two streams should communicate with each other, and this is crucial to phase prediction. In addition, we propose frequency transformation blocks to catch long-range correlations along the frequency axis. Visualization shows that the learned transformation matrix implicitly captures the harmonic correlation, which has been proven to be helpful for T-F spectrogram reconstruction. With these two innovations, PHASEN acquires the ability to handle detailed phase patterns and to utilize harmonic patterns, getting 1.76dB SDR improvement on AVSpeech + AudioSet dataset. It also achieves significant gains over Google's network on this dataset. On Voice Bank + DEMAND dataset, PHASEN outperforms previous methods by a large margin on four metrics.

Camera trace is a unique noise produced in digital imaging process. Most existing forensic methods analyze camera trace to identify image origins. In this paper, we address a new low-level vision problem, camera trace erasing, to reveal the weakness of trace-based forensic methods. A comprehensive investigation on existing anti-forensic methods reveals that it is non-trivial to effectively erase camera trace while avoiding the destruction of content signal. To reconcile these two demands, we propose Siamese Trace Erasing (SiamTE), in which a novel hybrid loss is designed on the basis of Siamese architecture for network training. Specifically, we propose embedded similarity, truncated fidelity, and cross identity to form the hybrid loss. Compared with existing anti-forensic methods, SiamTE has a clear advantage for camera trace erasing, which is demonstrated in three representative tasks.

State-of-the-art image classification algorithms building upon convolutional neural networks (CNNs) are commonly trained on large annotated datasets of high-quality images. When applied to low-quality images, they will suffer a significant degradation in performance, since the structural and statistical properties of pixels in the neighborhood are obstructed by image degradation. To address this problem, this paper proposes a novel deep degradation prior for low-quality image classification. It is based on statistical observations that, in the deep representation space, image patches with structural similarity have uniform distribution even if they come from different images, and the distributions of corresponding patches in low- and high-quality images have uniform margins under the same degradation condition. Therefore, we propose a feature de-drifting module (FDM) to learn the mapping relationship between deep representations of low- and high- quality images, and leverage it as a deep degradation prior (DDP) for low-quality image classification. Since the statistical properties are independent to image content, deep degradation prior can be learned on a training set of limited images without supervision of semantic labels and served in a form of “plugging-in” module of the existing classification networks to improve their performance on degraded images. Evaluations on the benchmark dataset ImageNet-C demonstrate that our proposed DDP can improve the accuracy of the pre-trained network model by more than 20% under various degradation conditions. Even under the extreme setting that only 10 images from CUB-C dataset are used for the training of DDP, our method improves the accuracy of VGG16 on ImageNet-C from 37% to 55%.

We consider the tracking problem as a special type of object detection problem, which we call instance detection. With proper initialization, a detector can be quickly converted into a tracker by learning the new instance from a single image. We find that model-agnostic meta-learning (MAML) offers a strategy to initialize the detector that satisfies our needs. We propose a principled three-step approach to build a high-performance tracker. First, pick any modern object detector trained with gradient descent. Second, conduct offline training (or initialization) with MAML. Third, perform domain adaptation using the initial frame. We follow this procedure to build two trackers, named Retina-MAML and FCOS-MAML, based on two modern detectors RetinaNet and FCOS. Evaluations on four benchmarks show that both trackers are competitive against state-of-the-art trackers. On OTB-100, Retina-MAML achieves the highest ever AUC of 0.712. On TrackingNet, FCOS-MAML ranks the first on the leader board with an AUC of 0.757 and the normalized precision of 0.822. Both trackers run in real-time at 40 FPS.

In this paper, we aim at solving a distributed machine learning problem under Byzantine attacks. In the distributed system, a number of workers (termed as Byzantine workers) could send arbitrary messages to the master and bias the learning process, due to data corruptions, computation errors or malicious attacks. Prior work has considered a total variation (TV) norm-penalized approximation formulation to handle Byzantine attacks, where the TV norm penalty forces the regular workers' local variables to be close, and meanwhile, tolerates the outliers sent by the Byzantine workers. The stochastic subgradient method, which does not consider the problem structure, is shown to be able to solve the TV norm-penalized approximation formulation. In this paper, we propose a stochastic alternating direction method of multipliers (ADMM) that utilizes the special structure of the TV norm penalty. The stochastic ADMM iterates are further simplified, such that the iteration-wise communication and computation costs are the same as those of the stochastic subgradient method. Numerical experiments on the COVERTYPE and MNIST dataset demonstrate the resilience of the proposed stochastic ADMM to various Byzantine attacks.

We propose an image coding scheme that compresses image into semantically scalable bitstream using deep neural networks. This scheme is expected to support intelligent analysis when the bitstream is partially decoded, as well as high-fidelity reconstruction of image when the bitstream is completely decoded. We implement such a semantically scalable image coding scheme based on semantic map. In the proposed scheme, the original image is firstly semantically segmented and the semantic map is compressed as the base layer. Then, the original image is segmented into several individual objects according to the semantic map, and each object is coded separately. A recurrent neural network-based encoder is used to compress these objects at several quality levels. At the decoder side, the semantic map can be directly applied for intelligent analysis. A generative adversarial network is used to synthesize a rough image using the semantic map. If user is interested in a certain object, more bits can be transmitted to enhance the quality of the object. Experimental results show that the proposed method achieves comparable compression performance with JPEG2000 at high bit rates, while facilitates intelligent analysis at low bit rates.

Video moment localization aims to localize a specific moment in a video by a natural language query. Previous works either use alignment information to find out the best-matching candidate (i.e., top-down approach) or use discrimination information to predict the temporal boundaries of the match (i.e., bottom-up approach). Little research has taken both the candidate-level alignment information and frame-level boundary information together and considers the complementarity between them. In this paper, we propose a unified top-down and bottom-up approach called Dual Path Interaction Network (DPIN), where the alignment and discrimination information are closely connected to jointly make the prediction. Our model includes a boundary prediction pathway encoding the frame-level representation and an alignment pathway extracting the candidate-level representation. The two branches of our network predict two complementary but different representations for moment localization. To enforce the consistency and strengthen the connection between the two representations, we propose a semantically conditioned interaction module. The experimental results on three popular benchmarks (i.e., TACoS, Charades-STA, and Activity-Caption) demonstrate that the proposed approach effectively localizes the relevant moment and outperforms the state-of-the-art approaches.

Recent image-specific Generative Adversarial Networks (GANs) provide a way to learn generative models from a single image instead of a large dataset. However, the semantic meaning of patches inside a single image is less explored. In this work, we first define the task of Semantic Image Analogy: given a source image and its segmentation map, along with another target segmentation map, synthesizing a new image that matches the appearance of the source image as well as the semantic layout of the target segmentation. To accomplish this task, we propose a novel method to model the patch-level correspondence between semantic layout and appearance of a single image by training a single-image GAN that takes semantic labels as conditional input. Once trained, a controllable redistribution of patches from the training image can be obtained by providing the expected semantic layout as spatial guidance. The proposed method contains three essential parts: 1) a self-supervised training framework, with a progressive data augmentation strategy and an alternating optimization procedure; 2) a semantic feature translation module that predicts transformation parameters in the image domain from the segmentation domain; and 3) a semantics-aware patch-wise loss that explicitly measures the similarity of two images in terms of patch distribution. Compared with existing solutions, our method generates much more realistic results given arbitrary semantic labels as conditional input.

In this paper, we propose a novel space-time video super-resolution method, which aims to recover a high-frame-rate and high-resolution video from its low-frame-rate and low-resolution observation. Existing solutions seldom consider the spatial-temporal correlation and the long-term temporal context simultaneously and thus are limited in the restoration performance. Inspired by the epipolar-plane image used in multi-view computer vision tasks, we first propose the concept of temporal-profile super-resolution to directly exploit the spatial-temporal correlation in the long-term temporal context. Then, we specifically design a feature shuffling module for spatial retargeting and spatial-temporal information fusion, which is followed by a refining module for artifacts alleviation and detail enhancement. Different from existing solutions, our method does not require any explicit or implicit motion estimation, making it lightweight and flexible to handle any number of input frames. Comprehensive experimental results demonstrate that our method not only generates superior space-time video super-resolution results but also retains competitive implementation efficiency.

The amount of volumetric brain image increases rapidly, which requires a vast amount of resources for storage and transmission, so it's urgent to explore an efficient volumetric compression method. Recent years have witnessed the progress of deep learning-based approaches for two-dimensional (2D) natural image compression, but the field of learned volumetric image compression still remains unexplored. In this paper, we propose the first end-to-end learning framework for volumetric image compression by extending the advanced techniques of 2D image compression to volumetric images. Specifically, a convolutional autoencoder is used to compress 3D image cubes, and the non-local attention models are embedded in the convolutional autoencoder to jointly capture local and global correlations. Both hyperprior and autoregressive models are used to perform the conditional probability estimation in entropy coding. To reduce model complexity, we introduce a convolutional long short-term memory network for the autoregressive model based on channel-wise prediction. Experimental results on volumetric mouse brain images show that the proposed method outperforms JPEG2000-3D, HEVC and state-of-the-art 2D methods.

We propose a Deep Boosting Framework (DBF) for real-world image denoising by integrating the deep learning technique into the boosting algorithm. The DBF replaces conventional handcrafted boosting units by elaborate convolutional neural networks, which brings notable advantages in terms of both performance and speed. We design a lightweight Dense Dilated Fusion Network (DDFN) as an embodiment of the boosting unit, which addresses the vanishing of gradients during training due to the cascading of networks while promoting the efficiency of limited parameters. The capabilities of the proposed method are first validated on several representative simulation tasks including non-blind and blind Gaussian denoising and JPEG image deblocking. We then focus on a practical scenario to tackle with the complex and challenging real-world noise. To facilitate leaning-based methods including ours, we build a new Real-world Image Denoising (RID) dataset, which contains 200 pairs of high-resolution images with diverse scene content under various shooting conditions. Moreover, we conduct comprehensive analysis on the domain shift issue for real-world denoising and propose an effective one-shot domain transfer scheme to address this issue. Comprehensive experiments on widely used benchmarks demonstrate that the proposed method significantly surpasses existing methods on the task of real-world image denoising. Code and dataset are available at https://github.com/ngchc/deepBoosting.

Light field images taken by plenoptic cameras often have a tradeoff between spatial and angular resolutions. In this paper, we propose a novel spatial super-resolution approach for light field images by jointly exploiting internal and external similarities. The internal similarity refers to the correlations across the angular dimensions of the 4D light field itself, while the external similarity refers to the cross-scale correlations learned from an external light field dataset. Specifically, we advance the classic projection-based method that exploits the internal similarity by introducing the intensity consistency checking criterion and a back-projection refinement, while the external correlation is learned by a CNN-based method which aggregates all warped high-resolution sub-aperture images upsampled from the low-resolution input using a single image super-resolution method. By analyzing the error distributions of the above two methods and investigating the upperbound of combining them, we find that the internal and external similarities are complementary to each other. Accordingly, we further propose a pixel-wise adaptive fusion network to take advantage of both their merits by learning a weighting matrix. Experimental results on both synthetic and real-world light field datasets validate the superior performance of the proposed approach over the state-of-the-arts.

Liver cancer is a leading cause of cancer deaths worldwide due to its high morbidity and mortality. Histopathological image analysis (HIA) is a crucial step in the early diagnosis of liver cancer and is routinely performed manually. However, this process is time-consuming, error-prone, and easily affected by the expertise of pathologists. Recently, computer-aided methods have been widely applied to medical image analysis; however, the current medical image analysis studies have not yet focused on the histopathological morphology of liver cancer due to its complex features and the insufficiency of training images with detailed annotations. This paper proposes a deep learning method for liver cancer histopathological image classification using only global labels. To compensate for the lack of detailed cancer region annotations in those images, patch features are extracted and fully utilized. Transfer learning is used to obtain the patch-level features and then combined with multiple-instance learning to acquire the image-level features for classification. The method proposed here solves the processing of large-scale images and training sample insufficiency in liver cancer histopathological images for image classification. The proposed method can distinguish and classify liver histopathological images as abnormal or normal with high accuracy, thus providing support for the early diagnosis of liver cancer.

Reconstruction of neuronal populations from ultra-scale optical microscopy (OM) images is essential to investigate neuronal circuits and brain mechanisms. The noises, low contrast, huge memory requirement, and high computational cost pose significant challenges in the neuronal population reconstruction. Recently, many studies have been conducted to extract neuron signals using deep neural networks (DNNs). However, training such DNNs usually relies on a huge amount of voxel-wise annotations in OM images, which are expensive in terms of both finance and labor. In this paper, we propose a novel framework for dense neuronal population reconstruction from ultra-scale images. To solve the problem of high cost in obtaining manual annotations for training DNNs, we propose a progressive learning scheme for neuronal population reconstruction (PLNPR) which does not require any manual annotations. Our PLNPR scheme consists of a traditional neuron tracing module and a deep segmentation network that mutually complement and progressively promote each other. To reconstruct dense neuronal populations from a terabyte-sized ultra-scale image, we introduce an automatic framework which adaptively traces neurons block by block and fuses fragmented neurites in overlapped regions continuously and smoothly. We build a dataset "VISoR-40" which consists of 40 large-scale OM image blocks from cortical regions of a mouse. Extensive experimental results on our VISoR-40 dataset and the public BigNeuron dataset demonstrate the effectiveness and superiority of our method on neuronal population reconstruction and single neuron reconstruction. Furthermore, we successfully apply our method to reconstruct dense neuronal populations from an ultra-scale mouse brain slice. The proposed adaptive block propagation and fusion strategies greatly improve the completeness of neurites in dense neuronal population reconstruction.

In this article, we introduce an image/video restoration approach by utilizing the high-dimensional similarity in images/videos. After grouping similar patches from neighboring frames, we propose to build a multiplanar autoregressive (AR) model to exploit the correlation in cross-dimensional planes of the patch group, which has long been neglected by previous AR models. To further utilize the nonlocal self-similarity in images/videos, a joint multiplanar AR and low-rank based approach is proposed (MARLow) to reconstruct patch groups more effectively. Moreover, for video restoration, the temporal smoothness of the restored video is constrained by the Markov random field (MRF), where MRF encodes a priori knowledge about consistency of patches from neighboring frames. Specifically, we treat different restoration results (from different patch groups) of a certain patch as labels of an MRF, and temporal consistency among these restored patches is imposed. The proposed method is also suitable for other restoration applications such as interpolation and text removal. Extensive experimental results demonstrate that the proposed approach obtains encouraging performance comparing with state-of-the-art methods.

Existing methods for single image super-resolution (SR) are typically evaluated with synthetic degradation models such as bicubic or Gaussian downsampling. In this paper, we investigate SR from the perspective of camera lenses, named as CameraSR, which aims to alleviate the intrinsic tradeoff between resolution (R) and field-of-view (V) in realistic imaging systems. Specifically, we view the R-V degradation as a latent model in the SR process and learn to reverse it with realistic low- and high-resolution image pairs. To obtain the paired images, we propose two novel data acquisition strategies for two representative imaging systems (i.e., DSLR and smartphone cameras), respectively. Based on the obtained City100 dataset, we quantitatively analyze the performance of commonly-used synthetic degradation models, and demonstrate the superiority of CameraSR as a practical solution to boost the performance of existing SR methods. Moreover, CameraSR can be readily generalized to different content and devices, which serves as an advanced digital zoom tool in realistic imaging systems.

The goal of multi-domain learning is to improve the performance over multiple domains by making full use of all training data from them. However, variations of feature distributions across different domains result in a non-trivial solution of multi-domain learning. The state-of-the-art work regarding multi-domain classification aims to extract domain-invariant features and domain-specific features independently. However, they view the distributions of features from different classes as a general distribution and try to match these distributions across domains, which lead to the mixture of features from different classes across domains and degrade the performance of classification. Additionally, existing works only force the shared features among domains to be orthogonal to the features in the domain-specific network. However, redundant features between the domain-specific networks still remain, which may shrink the discriminative ability of domain-specific features. Therefore, we propose an end-to-end network to obtain the more optimal features, which we call compact features. We propose to extract the domain-invariant features by matching the joint distributions of different domains, which have dis- tinct boundaries between different classes. Moreover, we add an orthogonal constraint between the private features across domains to ensure the discriminative ability of the domain-specific space. The proposed method is validated on three landmark datasets, and the results demonstrate the effectiveness of our method.

The greatest challenge facing visual object tracking is the simultaneous requirements on robustness and discrimination power. In this paper, we propose a SiamFC-based tracker, named SPM-Tracker, to tackle this challenge. The basic idea is to address the two requirements in two separate matching stages. Robustness is strengthened in the coarse matching (CM) stage through generalized training while discrimination power is enhanced in the fine matching (FM) stage through a distance learning network. The two stages are connected in series as the input proposals of the FM stage are generated by the CM stage. They are also connected in parallel as the matching scores and box location refinements are fused to generate the final results. This innovative series-parallel structure takes advantage of both stages and results in superior performance. The proposed SPM-Tracker, running at 120 fps on GPU, achieves an AUC of 0.687 on OTB-100 and an EAO of 0.434 on VOT-16, exceeding other real-time trackers by a notable margin.

In this paper, we aim at completing a large-scale low-rank matrix over a distributed network, which is subject to Byzantine attacks. We consider solving a nonconvex matrix factorization model with the distributed successive over-relaxation (SOR) method, where the distributed workers compute their private matrices using their own training data and the public matrix sent by the master, while the master updates the public matrix through aggregating the private matrices sent by the workers. However, the Byzantine workers could deliberately send faulty messages to the master so as to bias the optimization process. To address this issue, we propose to replace the aggregation step in the distributed SOR method by several state-of-the-art robust ones: geometric median, median, Krum and h-Krum. We conduct numerical experiments on the Netflix dataset and demonstrate the effectiveness of the proposed robust aggregation strategies in handling Byzantine attacks.

We study the video super-resolution (SR) problem for facilitating video analytics tasks, e.g. action recognition, instead of for visual quality. The popular action recognition methods based on convolutional networks, exemplified by two-stream networks, are not directly applicable on video of low spatial resolution. This can be remedied by performing video SR prior to recognition, which motivates us to improve the SR procedure for recognition accuracy. Tailored for two-stream action recognition networks, we propose two video SR methods for the spatial and temporal streams respectively. On the one hand, we observe that regions with action are more important to recognition, and we propose an optical-flow guided weighted mean-squared-error loss for our spatial-oriented SR (SoSR) network to emphasize the reconstruction of moving objects. On the other hand, we observe that existing video SR methods incur temporal discontinuity between frames, which also worsens the recognition accuracy, and we propose a siamese network for our temporal-oriented SR (ToSR) training that emphasizes the temporal continuity between consecutive frames. We perform experiments using two state-of-the-art action recognition networks and two well-known datasets--UCF101 and HMDB51. Results demonstrate the effectiveness of our proposed SoSR and ToSR in improving recognition accuracy.

The rapid advancement of high-speed whole-brain imaging techniques poses great challenge for the existing image compression schemes. To accommodate the huge amount of data on limited storage and also facilitate transmission, high compression rates are required, leading to noticeable image degradations that hinder the subsequent scientific observation. In this paper, we propose a deep learning framework for the enhancement of compressed brain images, which learns an efficient post-processing network for the reduction of compression artifacts. This framework is designed for the 3D volumetric brain image with high dynamic range by introducing novel network components, yet it is also applicable to other types of biomedical images. Comprehensive experiments demonstrate the superiority of the proposed method over existing solutions. To the best of our knowledge, this is the first framework for 3D HDR brain image compression with deep enhancement.

Contrast enhancement and noise removal are coupled problems for low-light image enhancement. The existing Retinex based methods do not take the coupling relation into consideration, resulting in under or over-smoothing of the enhanced images. To address this issue, this paper presents a novel progressive Retinex framework, in which illumination and noise of low-light image are perceived in a mutually reinforced manner, leading to noise reduction low-light enhancement results. Specifically, two fully pointwise convolutional neural networks are devised to model the statistical regularities of ambient light and image noise respectively, and to leverage them as constraints to facilitate the mutual learning process. The proposed method not only suppresses the interference caused by the ambiguity between tiny textures and image noises, but also greatly improves the computational efficiency. Moreover, to solve the problem of insufficient training data, we propose an image synthesis strategy based on camera imaging model, which generates color images corrupted by illumination-dependent noises. Experimental results on both synthetic and real low-light images demonstrate the superiority of our proposed approaches against the State-Of-The-Art (SOTA) low-light enhancement methods.

This paper describes our solution for the Content-Based Video Relevance Prediction (CBVRP) challenge, where the task is to predict user click-through behavior on new TV series or new movies according to the user's historical behavior. We consider the task as a session-based recommendation problem and we focus on the modeling of the session. Thus, we use the Bidirectional Encoder Representations from Transformer (BERT) methodology and propose a BERT for session-based recommendation (BERT4SessRec) method. Our method has two stages: in the pre-training stage, we use all sessions as training data and train the bidirectional session encoder with the masking trick; in the fine-tuning stage, we use the provided click-through data and train the click-through prediction network. Our method achieves session representations with the help of BERT, which effectively captures the bidirectional correlation in each session. In addition, the pre-training stage makes full use of all sessions, overcoming the positive-negative imbalance problem of the click-through data. We report the results of using different kinds of features on the test set of the challenge, which verify the effectiveness of our method.

In this paper, we propose a novel hybrid network with Laplacian enhancing unit for image enhancement. We combine the merits of two representative enhancement methods, i.e., the scaling scheme and the generative scheme, by forming a hybrid enhancing module. Meanwhile, we model image enhancement in a progressive manner with a deep cascading CNN architecture, in which the previous feature maps are used to enhance subsequent features to get an improved performance. Specifically, we propose a Laplacian enhancing unit, which can adjustably enhance the detail information by adding the residual of previous feature maps. This unit is embedded across layers for progressively enhancing the features. We build our network on the U-Net architecture and name it Hybrid Progressive Enhancing U-Net. Experiments show that our method achieves superior image enhancement results compared with the state-of-the-arts, while retaining competitive implementation efficiency.

Signal degradation is ubiquitous, and computational restoration of degraded signal has been investigated for many years. Recently, it is reported that the capability of signal restoration is fundamentally limited by the so-called perception-distortion tradeoff, i.e. the distortion and the perceptual difference between the restored signal and the ideal "original" signal cannot be made both minimal simultaneously. Distortion corresponds to signal fidelity and perceptual difference corresponds to perceptual naturalness, both of which are important metrics in practice. Besides, there is another dimension worthy of consideration--the semantic quality of the restored signal, i.e. the utility of the signal for recognition purpose. In this paper, we extend the previous perception-distortion tradeoff to the case of classification-distortion-perception (CDP) tradeoff, where we introduced the classification error rate of the restored signal in addition to distortion and perceptual difference. In particular, we consider the classification error rate achieved on the restored signal using a predefined classifier as a representative metric for semantic quality. We rigorously prove the existence of the CDP tradeoff, i.e. the distortion, perceptual difference, and classification error rate cannot be made all minimal simultaneously. We also provide both simulation and experimental results to showcase the CDP tradeoff. Our findings can be useful especially for computer vision research where some low-level vision tasks (signal restoration) serve for high-level vision tasks (visual understanding). Our code and models have been published.

We propose a novel hybrid imaging system to acquire 4D high-speed hyperspectral (HSHS) videos with high spatial and spectral resolution. The proposed system consists of two branches: one branch performs Nyquist sampling in the temporal dimension while integrating the whole spectrum, resulting in a high-frame-rate panchromatic video; the other branch performs compressive sampling in the spectral dimension with longer exposures, resulting in a low-frame-rate hyperspectral video. Owing to the high light throughput and complementary sampling, these two branches jointly provide reliable measurements for recovering the underlying HSHS video. Moreover, the panchromatic video can be used to learn an over-complete 3D dictionary to represent each band-wise video sparsely, thanks to the inherent structural similarity in the spectral dimension. Based on the joint measurements and the self-adaptive dictionary, we further propose a simultaneous spectral sparse (3S) model to reinforce the structural similarity across different bands and develop an efficient computational reconstruction algorithm to recover the HSHS video. Both simulation and hardware experiments validate the effectiveness of the proposed approach. To the best of our knowledge, this is the first time that hyperspectral videos can be acquired at a frame rate up to 100fps with commodity optical elements and under ordinary indoor illumination.

This paper proposes an unsupervised CNN-based method for explicit depth estimation from light field, which learns an end-to-end mapping from a 4D light field to the corresponding disparity map without the supervision of groundtruth depth. Specifically, we design a combined loss function imposing both compliance and divergence constraints on the warped sub-aperture images to the central view, which guarantees our network to generate an accurate and robust disparity map. Furthermore, we find that increasing the number of referenced views in depth feature extraction and complementing missing information caused by warping greatly boost the performance of our network. Due to the difficulty of obtaining groundtruth depth of real-world scenes in practice, the proposed method is much more feasible than supervised learning. On the other hand, compared with traditional non-learning methods, the proposed method better exploits the correlations in the 4D light field and generates superior depth results both quantitatively and qualitatively. Also, the proposed method helps improve the performance of subsequent applications based on the estimated depth, e.g., spatial super-resolution of light field.

This paper proposes an image upsampling scheme by introducing a new gradient field sharpening transform that converts the blurry gradient field of upsampled low-resolution (LR) image to a much sharper gradient field of original high-resolution (HR) image. Different from the existing methods that need to figure out the whole gradient profile structure and locate the edge points, we derive a new approach that sharpens the gradient field adaptively only based on the pixels in a small neighborhood. To maintain image contrast, image gradient is adaptively scaled to keep the integral of gradient field stable. Finally the HR image is reconstructed by fusing the LR image with the sharpened HR gradient field. Experimental results demonstrate that the proposed algorithm can generate more accurate gradient field and produce super-resolved images with better objective and visual qualities. Another advantage is that the proposed gradient sharpening transform is very fast and suitable for low-complexity applications.

For video action recognition, convolutional neural networks (CNNs) especially two-stream CNNs have achieved remarkable progress in the recent years. However, most of the CNNs for action recognition are trained with high-resolution videos and not scale invariant, making it problematic to apply the trained CNNs directly on low-resolution videos. One possible solution to the problem is performing super-resolution (SR) prior to action recognition. In this paper, we investigate the effects of CNN-based video SR on the action recognition accuracy. We adopt a well trained two-stream CNN for action recognition, and analyze the spatial and temporal streams separately. For the spatial stream, we observe that video SR may improve the PSNR but may incur drop in recognition accuracy, this phenomenon is further analyzed in this paper. For the temporal stream, we observe that frame-by-frame SR may produce temporal inconsistency between consecutive video frames, which also incurs drop in recognition accuracy. We then propose a temporal consistency-oriented method for video SR, which indeed improves the recognition accuracy. Finally, we perform proper fusion of the two streams, and achieve a recognition accuracy of 88.95% on the UCF101 dataset when the input video is down-sampled by a factor of 4, compared to 93.49% accuracy on the original-resolution videos.

Neuron segmentation is an important task for automatic analyses of brain images that are of huge volume. Previous methods for neuron segmentation rely on handcrafted image features, and have difficulty in coping with high-resolution, low signal-to-noise-ratio brain images. Convolutional neural network (CNN) has achieved remarkable success in natural image segmentation, but CNN requires accurately labeled data for training that are difficult to achieve on brain images of huge volume. In this paper, we present a weakly supervised learning strategy to deal with the inaccurate training data problem, and thus adopt 3D CNN to perform automatic soma segmentation from brain images. We test our method on our own collected mouse brain images that are of single-neuron resolution, and results show that 3D CNN-based method outperforms the traditional methods by a significant margin.

Hyperspectral video acquisition usually requires high complexity hardware and reconstruction algorithms. In this paper, we propose a low complexity CNN-based method for hyperspectral reconstruction from ubiquitous RGB videos, which effectively exploits the temporal redundancies within RGB videos and generates high-quality hyperspectral output. Specifically, given an RGB video, we first design an efficient motion compensation network to align the RGB frames and reduce the large motion. Then, we design a temporal-adaptive fusion network to exploit the inter-frame correlation. The fusion network has the ability to determine the optimum temporal dependency within successive frames, which further promotes the hyperspectral reconstruction fidelity. Preliminary experimental results validate the superior performance of the proposed method over previous learning-based methods. To the best of our knowledge, this is the first time that RGB videos are utilized for hyperspectral reconstruction through deep learning.

Light field, a promising representation to describe the scene appearance, is susceptible to various noise due to the current sensor design. This paper proposes a novel tensor-based denoising method for the 4D light field that consists of two main steps. First, we generalize the intrinsic tensor sparsity measure to light field images by exploiting the nonlocal similarity across the spatial and angular dimensions. Second, we further exploit the spatial-angular correlation by integrating light field super-resolution into the denoising process to eliminate the sub-pixel misalignment of different views. After a back-projection from the refined high-resolution central view under an intensity consistency criteria, the denoising performance for the light field can be boosted. Experimental results validate the superior performance of the proposed method in terms of both PSNR and visual quality on the HCI light field dataset.

Detecting small objects is a challenging task. Existing CNN-based objection detection pipeline faces such a dilemma: using a high-resolution image as input incurs high computational cost, but using a low-resolution image as input loses the feature representation of small objects and therefore leads to low accuracy. In this work, we propose a cascade mask generation framework to tackle this issue. The proposed framework takes in multi-scale images as input and processes them in ascending order of the scale. Each processing stage outputs object proposals as well as a region-of-interest (RoI) mask for the next stage. With RoI convolution, the masked regions can be excluded from the computation in the next stage. The procedure continues until the largest scale image is processed. Finally, the object proposals generated from multiple scales are classified by a post classifier. Extensive experiments on Tsinghua-Tencent 100K traffic sign benchmark demonstrate that our approach achieves state-of-the-art small object detection performance at a significantly improved speed-accuracy tradeoff compared with previous methods.

Micro-video sharing gains great popularity in recent years, which calls for effective recommendation algorithm to help user find their interested micro-videos. Compared with traditional online (e.g. YouTube) videos, micro-videos contributed by grass-root users and taken by smartphones are much shorter (tens of seconds) and more short of tags or descriptive text, making the recommendation of micro-videos a challenging task. In this paper, we investigate how to model user's historical behaviors so as to predict the user's click-through of micro-videos. Inspired by the recent deep network-based methods, we propose a Temporal Hierarchical Attention at Category- and Item-Level (THACIL) network for user behavior modeling. First, we use temporal windows to capture the short-term dynamics of user interests; Second, we leverage a category-level attention mechanism to characterize user's diverse interests, as well as an item-level attention mechanism for fine-grained profiling of user interests; Third, we adopt forward multi-head self-attention to capture the long-term correlation within user behaviors. Our proposed THACIL network was tested on MicroVideo-1.7M, a new dataset of 1.7 million micro-videos, coming from real data of a micro-video sharing service in China. Experimental results demonstrate the effectiveness of the proposed method in comparison with the state-of-the-art solutions.

Person re-identification aims to identify the same pedestrian across non-overlapping camera views. Deep learning techniques have been applied for person re-identification recently, towards learning representation of pedestrian appearance. This paper presents a novel Contextual-Attentional Attribute-Appearance Network ($\rm CA^3Net$) for person re-identification. The $\rm CA^3Net$ simultaneously exploits the complementarity between semantic attributes and visual appearance, the semantic context among attributes, visual attention on attributes as well as spatial dependencies among body parts, leading to discriminative and robust pedestrian representation. Specifically, an attribute network within $\rm CA^3Net$ is designed with an Attention-LSTM module. It concentrates the network on latent image regions related to each attribute as well as exploits the semantic context among attributes by a LSTM module. An appearance network is developed to learn appearance features from the full body, horizontal and vertical body parts of pedestrians with spatial dependencies among body parts. The $\rm CA^3Net$ jointly learns the attribute and appearance features in a multi-task learning manner, generating comprehensive representation of pedestrians. Extensive experiments on two challenging benchmarks, i.e., Market-1501 and DukeMTMC-reID datasets, have demonstrated the effectiveness of the proposed approach.

This letter presents a novel approach for simultaneous depth and spectral imaging with a cross-modal stereo system. Two images of the target scene are captured at the same time: one compressively sampled hyperspectral measurement and one panchromatic measurement. The underlying hyperspectral cube is first reconstructed by leveraging the compressive sensing theory, during which a self-adaptive dictionary is learned from the panchromatic measurement to facilitate the reconstruction. The depth information of the scene is then recovered by estimating a disparity map between the hyperspectral cube and the panchromatic measurement through stereo matching. This disparity map, once obtained, is used to align the hyperspectral and panchromatic measurements to boost the hyperspectral reconstruction in an iterative manner. Through hardware experiments, for the first time to our knowledge, we demonstrate a snapshot system that allows for simultaneous depth and spectral imaging. The proposed system is capable of recording depth and spectral videos of dynamic scenes.

This paper proposes a single-image super-resolution scheme by introducing a gradient field sharpening transform that converts the blurry gradient field of upsampled low-resolution (LR) image to a much sharper gradient field of original high-resolution (HR) image. Different from the existing methods that need to figure out the whole gradient profile structure and locate the edge points, we derive a new approach that sharpens the gradient field adaptively only based on the pixels in a small neighborhood. To maintain image contrast, image gradient is adaptively scaled to keep the integral of gradient field stable. Finally, the HR image is reconstructed by fusing the LR image with the sharpened HR gradient field. Experimental results demonstrate that the proposed algorithm can generate more accurate gradient field and produce super-resolved images with better objective and visual qualities. Another advantage is that the proposed gradient sharpening transform is very fast and suitable for low-complexity applications.

This paper presents the first snapshot hyperspectral light field imager in practice. Specifically, we design a novel hybrid camera system to obtain two complementary measurements that sample the angular and spectral dimensions respectively. To recover the full 5D hyperspectral light field from the severely undersampled measurements, we then propose an efficient computational reconstruction algorithm by exploiting the large correlations across the angular and spectral dimensions through self-learned dictionaries. Simulation on an elaborate hyperspectral light field dataset validates the effectiveness of the proposed approach. Hardware experimental results demonstrate that, for the first time to our knowledge, a 5D hyperspectral light field containing 9x9 angular views and 27 spectral bands can be acquired in a single shot.

In this paper, we investigate a convolutional neural network (CNN) approach for light field (LF) super-resolution (SR). We are motivated by the assumption that image priors can be embedded into CNN, and both external and internal correlations are important in LFSR. The LF images are indeed natural images except for its angular resolution, so the external correlations help to super-resolve a single image from a collection of general images, whilst the internal correlations are essential to enhance a single view in LF with the details in the other views. Accordingly, we propose a two-stage CNN, where the two stages exploit the external and internal correlations, respectively. Moreover, to improve the generalization ability of the second-stage CNN for inter-view SR, we propose to align different views at patch level to compensate for the disparity that is essential to LFSR, thus the second stage is termed multi-patch fusion CNN. Experimental results demonstrate the superior performance of our two-stage CNN compared with the state-of-the-art CNN-based SR methods.

This paper presents a novel super-resolution method for light field images by jointly exploiting internal and external similarities. The internal similarity refers to the correlations that exist across the angular dimensions of the 4D light field itself, while the external similarity refers to the correlations learned from a conventional 2D image dataset. Our key observation is that the internal and external similarities are complementary to each other, and we propose a depth-adaptive fusion scheme to take advantage of both their merits. Moreover, we improve the traditional projection-based method that exploits the internal similarity, by introducing a back-projection refinement and getting rid of the dependency on camera parameters. Experimental results on a variety of light field images validate the superior performance of the proposed method.

Light field (LF) cameras offer the capability of depth estimation in a single shot, which facilitates real-time 3D reconstruction of dynamic scenes. However, the accuracy of depth estimated from LF is still limited. Different from previous methods that generally focus on improving the fidelity of the central view depth, we argue that depth maps obtained at different views contain complementary information. Inspired by the principle of Kinect-fusion, we then propose a novel method for dense and accurate 3D reconstruction from LF images, namely, LF-fusion. Specifically, we use the iterative closest point (ICP) algorithm to register the point clouds generated from different views, and then employ a volumetric integration algorithm based on the truncated signed distance function (TSDF) to reconstruct the final 3D surface. Experiments demonstrate that the proposed method produces superior 3D reconstruction results on two representative LF datasets.

In this paper, we present an edge-guided generative adversarial network (EGGAN) for edge-based image inpainting that can be adopted in image compression and transmission error concealment. Our key idea is to integrate edges into the generative network, and train the generative network to minimize both gradient loss and adversarial loss. Given a corrupted image and the estimated edges of the missing area, the trained generative network is capable in generating the missing area in a visually plausible manner, and meanwhile reproducing the given edges faithfully. Experimental results on the challenging face images have shown the effectiveness of EGGAN.

Since low-resolution images may hamper the performance of optical character recognition (OCR), text image super-resolution (SR) has become an increasingly important problem in computer vision. Convolutional neural network (CNN) has been proposed for generic image SR as well as text image SR, but the previous works concern more on the objective quality (e.g. PSNR) rather than the OCR performance. In this paper, we propose a new loss function when training CNN for text image SR to facilitate OCR, and conduct model combination to further improve the performance. Also, we propose a simple yet effective image padding method to refine the image boundaries during SR. Experimental results show that we achieve an OCR accuracy of 78.10% on the ICDAR 2015 TextSR dataset, which is comparable with that of using the original high-resolution images (78.80%), and also exceeds the state-of-the-arts.

A light field consists of multiple views of a scene, which can be arranged and encoded like a pseudo sequence. Since the correlations between views are not equal and indeed content dependent, a well-constructed coding order and adaptive prediction structure will improve performance. In this paper, we propose an adaptive prediction structure for light field compression. While the coding order is inherited from the 2-D hierarchical coding order, the prediction structure is determined by the differences between scale-invariant feature transform (SIFT) descriptors of the views. Experimental results show that the proposed method leads to on average 5.71% BD-rate reduction compared with fixed prediction structure.

Leveraging the compressive sensing (CS) theory, coded aperture snapshot spectral imaging (CASSI) provides an efficient solution to recover 3D hyperspectral data from a 2D measurement. The dual-camera design of CASSI, by adding an uncoded panchromatic measurement, enhances the reconstruction fidelity while maintaining the snapshot advantage. In this paper, we propose an adaptive nonlocal sparse representation (ANSR) model to boost the performance of dual-camera compressive hyperspectral imaging (DCCHI). Specifically, the CS reconstruction problem is formulated as a 3D cube based sparse representation to make full use of the nonlocal similarity in both the spatial and spectral domains. Our key observation is that, the panchromatic image, besides playing the role of direct measurement, can be further exploited to help the nonlocal similarity estimation. Therefore, we design a joint similarity metric by adaptively combining the internal similarity within the reconstructed hyperspectral image and the external similarity within the panchromatic image. In this way, the fidelity of CS reconstruction is greatly enhanced. Both simulation and hardware experimental results show significant improvement of the proposed method over the state-of-the-art.

In this paper, we propose a novel and practical system for robust online face tracking in surveillance videos. The proposed system has two contributions: 1) sustained high performance for long-term tracking even when faces come in and out of the view frequently, and 2) extremely low complexity which allows for real-time deployment on various platforms. These advantages are achieved by designing a regular update framework based on a state-of-the-art face detector and a new histogram-assisted KLT (HAKLT) tracker. Experimental results demonstrate a superior and super-fast (>100fps) practical face tracking system.

Coded aperture snapshot spectral imaging (CASSI) has been demonstrated as a feasible solution to recover a 3D hyperspectral image by using a single 2D measurement. In this paper, we propose a new hybrid camera design for CASSI to capture high quality hyperspectral images while maintaining the snapshot advantage. Specifically, we employ a complementary RGB camera in conjunction with the CASSI system. The recorded RGB image can provide reliable spectral clue of the scene. By combining the coded hyperspectral information from the CASSI branch and the uncoded color information from the RGB branch, hyperspectral images can be reconstructed with high fidelity. Furthermore, by conducting demosaicing on the raw RGB image as a preprocessing procedure, even better performance can be achieved. Both theoretical analysis and simulation results show improved accuracy of the proposed method compared to the state-of-the-arts.

The Kinect depth sensor has become a versatile tool in different research fields due to its low cost and ease of use, yet its implementation still remains a black box to a large extent. Moreover, the accuracy and speed of Kinect are limited for certain applications. To boost the performance of Kinect while preserving its advantages, we have proposed a depth acquisition method that uses speckle patterns generated by commodity laser emitters as in Kinect, but the speckle density is modulated to allow higher accuracy depth measurement. In this paper, we present the field-programmable gate array (FPGA) design and implementation for the above Kinectlike depth sensing. To achieve a high-speed performance on a normal FPGA (Altera Stratix III EP3SL150), we elaborate the hardware architecture to realize a full-pipeline framework for the two main modules: 1) pixel-wise phase generation and 2) normalized cross-correlation-based phase unwrapping. Experiments show a reconstruction speed of 572 fps at a resolution of 640 × 480 with improved depth accuracy over Kinect and that an even higher resolution is supported. The previous pattern design and the FPGA design in this paper jointly demonstrate the feasibility of next-generation fast, accurate, and low-cost depth sensors.

We propose a novel dual-camera design to acquire 4D high-speed hyperspectral (HSHS) videos with high spatial and spectral resolution. Our work has two key technical contributions. First, we build a dual-camera system that simultaneously captures a panchromatic video at a high frame rate and a hyperspectral video at a low frame rate, which jointly provide reliable projections for the underlying HSHS video. Second, we exploit the panchromatic video to learn an over-complete 3D dictionary to represent each band-wise video sparsely, and a robust computational reconstruction is then employed to recover the HSHS video based on the joint videos and the self-learned dictionary. Experimental results demonstrate that, for the first time to our knowledge, the hyperspectral video frame rate reaches up to 100fps with decent quality, even when the incident light is not strong.

Depth sensors based on Time-of-Flight (ToF) and Phase Shifting (PS) have complementary strengths and weaknesses. ToF can provide real-time depth but limited in resolution and sensitive to noise. PS can generate accurate and robust depth with high resolution but requires a number of patterns that leads to high latency. In this paper, we propose a novel fusion framework to take advantages of both ToF and PS. The basic idea is using the coarse depth from ToF to disambiguate the wrapped depth from PS. Specifically, we address two key technical problems: cross-modal calibration and interference-free synchronization between ToF and PS sensors. Experiments demonstrate that the proposed method generates accurate and robust depth with high resolution and low latency, which is beneficial to tremendous applications.

Time multiplexing (TM) and spatial neighborhood (SN) are two mainstream structured light techniques widely used for depth sensing. The former is well known for its high accuracy and the latter for its low delay. In this paper, we explore a new paradigm of scalable depth sensing to integrate the advantages of both the TM and SN methods. Our contribution is twofold. First, we design a set of hybrid structured light patterns composed of phase-shifted fringe and pseudo-random speckle. Under the illumination of the hybrid patterns, depth can be decently reconstructed either from a few consecutive frames with the TM principle for static scenes or from a single frame with the SN principle for dynamic scenes. Second, we propose a scene-adaptive depth sensing framework based on which a global or region-wise optimal depth map can be generated through motion detection. To validate the proposed scalable paradigm, we develop a real-time (20 fps) depth sensing system. Experimental results demonstrate that our method achieves an efficient balance between accuracy and speed during depth sensing that has rarely been exploited before.

This paper proposes novel density modulated binary patterns for depth acquisition. Similar to Kinect, the illumination patterns do not need a projector for generation and can be emitted by infrared lasers and diffraction gratings. Our key idea is to use the density of light spots in the patterns to carry phase information. Two technical problems are addressed here. First, we propose an algorithm to design the patterns to carry more phase information without compromising the depth reconstruction from a single captured image as with Kinect. Second, since the carried phase is not strictly sinusoidal, the depth reconstructed from the phase contains a systematic error. We further propose a pixel-based phase matching algorithm to reduce the error. Experimental results show that the depth quality can be greatly improved using the phase carried by the density of light spots. Furthermore, our scheme can achieve 20 fps depth reconstruction with GPU assistance.

In this paper, we present a novel single-shot method for dense and accurate 3D scanning. Our method takes advantage of two conventional techniques, i.e., stereo and Fourier fringe analysis (FFA). While FFA is competent for high-density and high-precision phase measurement, stereo solves the phase ambiguity caused by the periodicity of the fringe. By jointly using the intensity images and unwrapped phase maps from stereo, the pixel-wise absolute depth can be obtained through a sparse matching process efficiently and reliably. Due to its single-shot property and low complexity, the proposed method facilitates dense and accurate 3D scanning in time-critical applications.

Phase shifting is a widely used method for accurate and dense 3D reconstruction. However, at least three images of the same scene are required for each reconstruction, so measurement errors are inevitable in dynamic scenes, even with high-speed hardware. In this paper, we propose a Fourier transform assisted phase shifting method to overcome the motion vulnerability in phase shifting. A new model with motion-related phase shifts is formulated, and the coarse phase measurements obtained by Fourier transform profilemetry are used to estimate the unknown phase shifts. The phase errors caused by motion are greatly reduced in this way. Experimental results show that the proposed method can obtain accurate and dense 3D reconstruction of dynamic scenes, with regard to different kinds of motion.

The one-to-one correspondence between co-occurrence image patches of two different resolutions is extensively used in example-based super-resolution (SR). Due to the dimensionality gap between low resolution (LR) and high resolution (HR) spaces, however, an LR patch may correspond to a number of HR patches in practice. This ambiguity is difficult to be overcome with examples representing a deterministic mapping. In this paper, we propose a statistical method for exploiting the one-to-many correspondence between LR and HR patches, which we call soft information and decision. Soft information means an LR patch is mapped to a pixel-wise distribution of all its possible HR counterparts, rather than a single or a limited set of HR candidates. Relying on the soft information, example-based SR is then regarded as an optimization problem to best preserve the local consistency in the recovered HR image. This problem is solved with an efficient message passing algorithm with a factor graph model. The final decision on the HR pixel value is made upon the maximum a posteriori estimation and is called a soft decision. Experimental results demonstrate the superiority of the proposed method compared with the state-of-the-art methods, in terms of both the subjective and objective quality of synthesized HR images.

Time multiplexing and spatial neighborhood are two mainstream structured light techniques widely used for 3D shape measurement. In this paper, we explore a way to subtly integrate their advantages for scalable depth sensing. This is realized through a set of elaborate hybrid structured patterns, which consists of three sinusoidal fringe patterns with different initial phases modulated by a pseudo-random speckle signal. For temporally static scenes, a high resolution, high accuracy depth map can be recovered from the latest three frames by the phase-shifting method; for dynamic scenes, a decent depth map can still be recovered from the current single frame by image matching. Since it provides seamless transition between high quality and quick response options, our method validates a new paradigm of scalable depth sensing in practice.

Automatic focus and exposure are the key components in digital cameras nowadays, which jointly play an essential role for capturing a high quality image. In this paper, we make an attempt to address these two challenging issues for future depth cameras. Relying on a programmable projector, we establish a structured light system for depth sensing with focus and exposure adaptation. The basic idea is to change current illumination pattern and intensity locally according to the prior depth information. Consequently, object surfaces appearing at different depths in the scene can receive proper illumination respectively. In this way, more flexible and robust depth sensing can be achieved in comparison with fixed illumination, especially at near depth.

This correspondence presents an image compression approach that integrates our proposed parameter-assistant inpainting (PAI) to exploit visual redundancy in color images. In this scheme, we study different distributions of image regions and represent them with a model class. Based on that, an input image at the encoder side is divided into featured and non-featured regions at block level. The featured blocks fitting the predefined model class are coded by a few parameters, whereas the non-featured blocks are coded traditionally. At the decoder side, the featured regions are restored through PAI relying on both delivered parameters and surrounding information. Experimental results show that our method outperforms JPEG in featured regions by an average bit-rate saving of 76% at similar perceptual quality levels.

This paper proposes a robust single-image super-resolution method for enlarging low quality web image/video degraded by downsampling and compression. To simultaneously improve the resolution and perceptual quality of such web image/video, we bring forward a practical solution which combines adaptive regularization and learning-based super-resolution. The contribution of this work is twofold. First, we propose to analyze the image energy change characteristics during the iterative regularization process, i.e., the energy change ratio between primitive (e.g., edges, ridges and corners) and nonprimitive fields. Based on the revealed convergence property of the energy change ratio, appropriate regularization strength can then be determined to well balance compression artifacts removal and primitive components preservation. Second, we verify that this adaptive regularization can steadily and greatly improve the pair matching accuracy in learning-based super-resolution. Consequently, their combination effectively eliminates the quantization noise and meanwhile faithfully compensates the missing high-frequency details, yielding robust super-resolution performance in the compression scenario. Experimental results demonstrate that our solution produces visually pleasing enlargements for various web images/videos.

Example-based super-resolution recovers missing high frequencies in a magnified image by learning the correspondence between co-occurrence examples at two different resolution levels. As high-resolution examples usually contain more details and are of higher dimensionality in comparison with low-resolution ones, the mapping from low-resolution to high-resolution is an ill-posed problem. Rather than imposing more complicated mapping constraints, we propose to improve the mapping accuracy by enhancing low-resolution examples in terms of mapped features, e.g., derivatives and primitives. A feature enhancement method is presented through a combination of interpolation with prefiltering and non-blind sparse prior deblurring. By enhancing low-resolution examples, unique feature information carried by high-resolution examples is decreased. This regularization reduces the intrinsic dimensionality disparity between two different resolution examples and thus improves the feature mapping accuracy. Experiments demonstrate our super-resolution scheme with feature enhancement produces high quality results both perceptually and quantitatively.

This paper addresses the super-resolution problem for low quality cartoon videos widely distributed on the web, which are generated by downsampling and compression from the sources. To effectively eliminate the compression artifacts and meanwhile preserve the visually salient primitive components (e.g., edges, ridges and corners), we propose an adaptive regularization method depending on the degradation grade of each frame, followed by learning-based pair matching to further enhance the primitives in the upsampled frames. In addition, temporal consistency is considered a directive constraint in both the regularization and enhancement processes. Experimental results demonstrate our solution achieves a good balance between artifacts removal and primitive enhancement, providing perceptually high quality super-resolution results for various web cartoon videos.

This paper presents a new directional image interpolator, aiming to increase image resolution with high perceptual quality and low computational complexity. In our method, missing pixels in a magnified image are generated through linear interpolation on certain fixed supports to facilitate fast implementation, while local directional features are imposed on the adaptive interpolation weights which are determined by the gradients diffused from the low resolution image. Afterwards, a novel difference projection strategy is proposed to enforce the continuity of the magnified image by reusing the directional interpolator. Experimental results show that our method outperforms conventional bicubic and some existing adaptive interpolators, in terms of both the perceptual and quantitative quality.

This paper proposes a single-image super-resolution scheme for enlarging low quality thumbnail images widely distributed on the web, which are often generated by downsampling plus compression. To obtain visually pleasurable high-resolution versions for this kind of low-resolution images, we first adopt a PDE-based image regularization technique to alleviate the compression noise in the distorted thumbnails, and then use learning-based pair matching to further enhance the high-frequency details in the upsampled images. Experimental results show that our solution achieves better visual quality for both offline and online test images, compared with traditional methods.

This paper carves out an image compression approach that integrates our parameter-assistant inpainting (PAI) technique to exploit the visual redundancy inherent in color-gradation image regions. In our scheme, an input image is first classified at block level according to the degree of edge content as well as chromatic variation in each block. An exemplar selection approach is then adopted to skip a majority of the gradation blocks during encoding. Only their positions and certain parameters extracted for condensed description are encoded along with the reserved blocks. At the decoder side, the skipped regions are recovered through image inpainting, relying on both the delivered parameters and reserved regions. Experimental results show that our proposed method outperforms baseline JPEG at color-gradation regions by nearly 80% bits-saving, at similar visual quality levels.