The application of animation models in facial video compression has yielded significant coding gains, particularly at ultra-low bitrates. Despite notable advancements, research on portrait video scenes, especially in half-body and full-body contexts, remains underexplored. The mapping motion features and dynamic regions is often imprecise, leading to inaccurate motion parameters. Additionally, the reconstruction of occluded background is frequently suboptimal, as occluded regions lack ground-truth data. To this end, we propose a learning-based portrait video compression (LPVC) framework for half-body and full-body portrait videos with static backgrounds. Specifically, we first design a keypoint detector driven by human parsing to integrate semantic properties into the animation model. This facilitates richer features and enhances the motion features for dynamic areas. We further devise a background incremental coding scheme to reconstruct high-quality occluded backgrounds. The scheme incorporates background occlusion calculation and compensation modules to process the newly revealed background pixels between successive frames, eliminating redundant background transmission. Finally, a spatio-temporal portrait background fusion generative adversarial network (SPBF-GAN) is devised to learn spatio-temporal differential representations of videos. Experimental results demonstrate that our proposed scheme achieves satisfying perceptual performance at ultra-low bitrates and exhibits higher semantic fidelity in semantic analysis tasks such as human parsing. Code is available at: https://github.com/Chen8023/LPVC .

Object detection in low-light scenes is a challenging but widely discussed topic in computer vision. A common approach in low-light object detection involves employing cascaded architectures to connect enhancement and detection networks. This strategy aims to bridge the gap between low-light and normal-light images in both the image and feature domains, enabling the effective application of existing object detection networks to low-light scenarios. However, the cascade architecture ignores the intrinsic connection between low-light image enhancement and object detection tasks. To address this problem, we use a bilevel learning architecture to bridge these two tasks. Specifically, we take the parameters of the enhancement network as the main optimization objective and the parameters of the object detection network as the learnable constraints. In addition, to improve the practicality, we propose a method based on implicit function theory to approximate the solution of the corresponding gradient. Numerous experiments have demonstrated that our method significantly improves enhancement quality and detection accuracy.

To address the dual challenges of inadequate deep semantic feature representation and limited data diversity in bi-temporal remote sensing change detection (RSCD), we propose a collaborative optimization framework (SRA-VFM) integrating customized data augmentation and hierarchical feature interpretation. SRA-VFM comprises three core components: slice-reassemble augmentation (SRA), a dual-stream feature encoding-decoding network, and a multi-task segmentation head. The SRA module synthesizes diverse training samples through random slicing and semantic reassembly while preserving local feature consistency. The dual-stream encoder, based on the FastSAM pre-trained model, incorporates a top-down feature adapter to align pre-trained features with remote sensing data distributions via cross-level fusion and low-dimensional semantic mapping. The bottom-up decoder leverages semantically aligned pyramid features for progressive upsampling, restoring high-resolution spatial details and enhancing multi-scale representation. The multi-task head jointly optimizes change region detection and edge refinement, accelerating convergence and improving boundary localization. Extensive experiments on 5 benchmark datasets (LEVIR-CD, WHU-CD, CLCD, S2Looking, SYSU-CD) demonstrate SRA-VFM’s superiority over state-of-the-art (SOTA) methods, achieving mF1/mIoU of 96.01%/92.55%, 97.13%/94.53%, 88.91%/81.21%, 83.13%/48.91%, and 88.36%/79.67% respectively. Code will be publicly available upon publication.

Abstract Despite the existence of various super‐resolution (SR) methods, most of them focus on designing models for specific upscaling factors rather than fully exploiting inter‐scale correlation to improve efficiency. In contrast, multi‐scale SR methods can effectively reduce the redundancy of network parameters by aggregating the feature extraction processes corresponding to multiple scales into a unified process. The aim of this study is to enhance the compactness and efficiency of the SR model. Thus, an efficient multi‐scale SR method called the diverse branch feature refinement network (DBFRN) is proposed. By decoupling the training process and inference process based on the idea of structural re‐parameterization, multi‐branch topology is adopted to enrich multi‐scale learning and merge branches to achieve efficient inference with equivalent effects. Specifically, two re‐parameterization strategies are designed and two corresponding feature refinement blocks for different feature levels in multi‐scale SR network. Extensive experiments demonstrate that the proposed multi‐scale SR method is effective and efficient, and it can outperform advanced single‐scale methods in terms of quantity and quality.

Abstract Compared with convolutional operation, non‐local operation can directly capture long‐range dependencies and thus has a larger receptive field. However, the computation and memory consumption of non‐local operation is much higher than convolutional operation, so it cannot be used repeatedly as a general component directly. In this paper, in order to balance the accuracy and computational complexity of non‐local enhancement, the non‐local operation is simplified based on local importance‐based pooling, which can dynamically extract discriminative features during the down‐sampling process by learning adaptive weights. Such simplified non‐local enhancement is able to prevent unacceptable computational consumption caused by directly processing the entire feature maps containing a large number of features. In order to verify the effectiveness of the proposed method, 2D and 3D feature extraction blocks are constructed based on the simplified non‐local operations, and they are stacked as feature extraction networks for space‐time video super‐resolution task, which aims to increase resolution in both time and space simultaneously. Extensive experiments demonstrate that the proposed simplified non‐local networks can effectively improve the performance of space‐time video super‐resolution task both quantitatively and qualitatively.

The demands of high-quality videos captured by camera become bigger due to the rapid development of pattern recognition and artificial intelligence. Video denoising is the key technology to obtain clear videos. However, the research on video denoising is far from enough now. In this paper, we propose a video denoising method based on convolutional neural network architecture to reduce the noise from the sensor system. We improve the loss function of noise estimation by imposing adaptive penalty on under-estimation error of noise level which makes our method perform robustly. Furthermore, we make use of multi-level features to guide the spatial denoising, where multilayer semantic information of the image is regarded as the perceptual loss. Instead of relying on Optical Flow solving the characterization of inter-frame information, we utilize U-Net-like structure to handle motion implicitly. It is less computationally expensive and avoids distortions caused by inaccurate flow and object occlusion. In order to locate temporal features and suppress useless information, the attention mechanism is introduced to the skip connections of the U-Net-like structure. Experimental results demonstrate that the proposed algorithm outputs more convincing results in both peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) indexes when processing Gaussian noise, synthetic real noise, and real noise compared with selected approaches.

Frame interpolation is one of the most challenging tasks in the video processing field. Recent advances have demonstrated that the deep learning‐based frame interpolation methods are promising. However, the experiments show that most existing deep learning‐based methods have the same problem as traditional methods. When these algorithms handle severe occlusions, they will produce distortions, especially around the motion boundaries. To better synthesise the image of the motion areas, the authors design a mask‐guided frame synthesis model, which consists of multiple components, based on deep convolutional neural networks. The proposed model first estimates the asymmetric bi‐directional optical flows from the intermediate frame to the input frames. Then it estimates the occlusion‐aware masks, which can compensate for the optical flow inaccuracy based on optical flows and correlation information. Finally, the warped frames are adaptively fused under the guidance of the masks to generate a high‐quality intermediate frame. Furthermore, to generate more realistic video frames, they train the network model with the pixel‐based loss and the feature‐based loss in a step‐by‐step way. In the experiment, they analyse the proposed model and compare it with the high‐performance methods, both qualitative and quantitative results show that their method performs better.

A computer vision method is presented for the mobile robot to find humans in scene. Face detection is used for confirming humans. In order to reduce regions of search, optical flow algorithm is used to segment the image in advance. Asymmetric problems in face detection are explained, and relative solutions are put forward by bootstrapping strategy and asymmetric adaboost algorithm. In addition, fisher discriminant analysis further improves the performance of face detection. Multi-view face models are trained to accommodate practical face detection application. At last, experiments demonstrate that our multi-view face detector achieves high detection accuracy and fast detection speed on both standard testing datasets and real-life images.

This paper suggests a new chaos-based color image cipher with an efficient substitution keystream generation strategy. The hyperchaotic Lü system and logistic map are employed to generate the permutation and substitution keystream sequences for image data scrambling and mixing. In the permutation stage, the positions of colored subpixels in the input image are scrambled using a pixel-swapping mechanism, which avoids two main problems encountered when using the discretized version of area-preserving chaotic maps. In the substitution stage, we introduce an efficient keystream generation method that can extract three keystream elements from the current state of the iterative logistic map. Compared with conventional method, the total number of iterations is reduced by 3 times. To ensure the robustness of the proposed scheme against chosen-plaintext attack, the current state of the logistic map is perturbed during each iteration and the disturbance value is determined by plain-pixel values. The mechanism of associating the keystream sequence with plain-image also helps accelerate the diffusion process and increase the degree of randomness of the keystream sequence. Experimental results demonstrate that the proposed scheme has a satisfactory level of security and outperforms the conventional schemes in terms of computational efficiency.

Audio oriented integrity protection should consider the characteristics of audio signals based on the combination of audio application scenarios. However, in the current popular network interaction environment, traditional verification based solutions can no longer work. A kind of integrity protection scheme for audio business should be redesigned in these new scenarios. In this context, a method of audio integrity protection based on relative phase (RP-AIP) is proposed. Through the design of integrity object (I.O.), the integrity of audio can be abstracted as the completeness and accuracy of it. The I.O. is bound to the audio signal in a uniformly and randomly embedded manner, and the embedding rules are controlled by the relative phase characteristic of the audio itself. The model and strategy of RP-AIP are illustrated, and the corresponding process and algorithm are demonstrated as well. Simulation experiments illustrate the feasibility of the proposed solution and indicate the superiority of its performance.

Cloud computing is changing the way of developing, deploying and managing applications. However, as typical Internet-based applications, cloud computing services lack carrier-grade signaling control mechanism and cannot guaranty Quality of Service (QoS), which have actually become technical barriers for telecom operator to provide commercial public cloud services. On the other hand, as the core signaling architecture of Next Generation Networking (NGN), IP Multimedia Subsystem (IMS) is facing the problem of the lack of innovative value-added services. This paper presents an architecture to support cloud computing services over IMS. In the proposed architecture, cloud services are regarded as the general IMS applications and then cloud clients are allowed to access cloud services under the control of Session Initiation Protocol (SIP) signaling and QoS mechanism of IMS. This paper introduces architecture overview and cloud service relevant functional components, and mainly discusses several key issues including cloud notification mechanism, QoS and charging control of IMS-based cloud computing services.