Benefiting from the powerful data learning and representation capabilities of neural networks, Learned Image Compression (LIC) methods have demonstrated better Rate-Distortion (RD) performance than traditional image compression frameworks. In this paper, we analyze the role of latent variables in image compression, both qualitatively and quantitatively. We then propose a latent variable compensation method to mitigate the loss introduced by quantization. We also introduce a regularization term for the latent variable mean square error into the loss function, providing more explicit guidance for the compression and reconstruction of the model’s internal latent variables. Additionally, we propose a noise compensation method that acts as a plug-and-play component to enhance reconstruction in lossy image compression without significant additional encoding or decoding time. We also present a data augmentation technique involving image inversion, which helps the training set conform to the symmetry inherent in probabilistic modeling for image compression tasks. Extensive experiments demonstrate that the proposed method enhances rate-distortion metrics and visual quality.

Current diffusion-based super-resolution methods for remote sensing images face two critical limitations: (1) excessive computational demands due to iterative sampling; and (2) semantic inconsistency in reconstructed images caused by stochastic denoising. To address these challenges, we propose SfmSR, a Semantics-guided Flow Matching model that achieves fast (1-5 steps) and accurate reconstruction through a novel two-stage framework. Stage 1 performs large-scale pre-training on diverse remote sensing data to learn robust multi-scale representations, while Stage 2 introduces semantic-guided fine-tuning, where extracted high-level features dynamically regulate the flow matching process. This dual-phase approach reduces sampling steps by two orders of magnitude while maintaining stability through deterministic ODE-based generation. Experimental results demonstrate that SfmSR outperforms existing methods on remote sensing image datasets such as Potsdam and Toronto. Moreover, SfmSR exhibits significant advantages in model complexity and inference efficiency, achieving fast inference with fewer parameters and lower memory usage, thus meeting real-time requirements in practical applications. Compared to state-of-the-art (SOTA) methods, SfmSR not only achieves superior reconstruction quality but also demonstrates clear advantages in inference speed and computational efficiency.

Underwater images usually exhibit severe color cast, hazy appearance, and/or dark regions because of the complex lighting absorption and scattering in water. How to increase the quality of these degraded underwater images has emerged as a key issue for various underwater application tasks. Recent efforts have been made to deal with single type degradation, however, it is still challenging to deal with multiple degradations that usually coexist in an underwater image with a general network. The degradations in underwater images can be divided into medium-agnostic (hazy or low-light which also encountered in in-air images) and medium-specific (color distortion caused by the specific light attenuation property in water) ones. According to this observation, this article proposes a cascaded multimodule underwater image enhancement (UIE) framework to address the coexisted multiple degradations. In the proposed framework, an in-air image enhancement module and a novel proposed adaptive color channel compensation network (AC3Net) are cascaded, in which the former focuses primarily on solving medium-agnostic degradations and the latter is for handling the medium-specific degradation. This framework has good flexibility by cascading different types of in-air image enhancement networks with AC3Net to achieve various UIE. The effectiveness of the proposed framework has been extensively validated on various degraded underwater images as well as different underwater visual perception tasks.

Object detection accuracy degrades seriously in visually degraded scenes. A natural solution is to first enhance the degraded image and then perform object detection. However, it is suboptimal and does not necessarily lead to the improvement of object detection due to the separation of the image enhancement and object detection tasks. To solve this problem, we propose an image enhancement guided object detection method, which refines the detection network with an additional enhancement branch in an end-to-end way. Specifically, the enhancement branch and detection branch are organized in a parallel way, and a feature guided module is designed to connect the two branches, which optimizes the shallow feature of the input image in the detection branch to be as consistent as possible with that of the enhanced image. As the enhancement branch is frozen during training, such a design plays a role in using the features of enhanced images to guide the learning of object detection branch, so as to make the learned detection branch being aware of both image quality and object detection. When testing, the enhancement branch and feature guided module are removed, and so no additional computation cost is introduced for detection. Extensive experimental results, on underwater, hazy, and low-light object detection datasets, demonstrate that the proposed method can improve the detection performance of popular detection networks (YOLO v3, Faster R-CNN, DetectoRS) significantly in visually degraded scenes.

Obtaining discriminative features is one of the key problems in three-dimensional (3D) shape retrieval. Recently, deep metric learning-based 3D shape retrieval methods have attracted the researchers’ attention and have achieved better performance. The triplet-center loss can learn more discriminative features than traditional classification loss, and it has been successfully used in deep metric learning-based 3D shape retrieval task. However, it has a hard margin parameter that only leverages part of the training data in each mini-batch. Moreover, the margin parameter is often determined by experience and remains unchanged during the training process. To overcome the above limitations, we propose the soft margin triplet-center loss, which replaces the margin with the nonparametric soft margin. Furthermore, we combined the proposed soft margin triplet-center loss with the softmax loss to improve the training efficiency and the retrieval performance. Extensive experimental results on two popular 3D shape retrieval datasets have validated the effectiveness of the soft margin triplet-center loss, and our proposed 3D shape retrieval method has achieved better performance than other state-of-the-art method.

Deep learning has been successfully applied to learn local feature descriptors in recent years. However, most of existing methods are supervised methods relying on a large number of labeled training patches, which are also proposed for learning real valued descriptors. In this paper, we propose a novel unsupervised deep learning method for binary descriptor learning. The binary descriptors are much more compact and efficient than the real valued descriptors and unsupervised leaning is highly required in many applications due to its label-free characteristic as the annotations are sometimes expensive to obtain. The core idea of our method is to explore the locality consistency in the descriptor space as well as to distinguish different patches while maintaining the ability to match a patch with its geometric transformed ones. We also give a theorical analysis about the role of batch normalization in learning effective binary descriptors. Benefited from this analysis, there is no need to append two additional losses on minimizing the quantization error and maximizing the entropy to the final learning objective like previous works did, thus simplifying our network training. Experiments on four benchmarks demonstrate that the proposed method is able to learn binary descriptors significantly outperforming previous unsupervised binary descriptors, even superior to most supervised ones. Especially, it obtains 21.2% of improvement on the UBC Phototour dataset, and 19.8%, 26.7%, 26.0% of improvements for patch verification, matching, retrieval tasks respectively on the HPatches dataset compared to the previous best unsupervised method.

This study introduces a novel multi‐feature‐based non‐rigid three‐dimensional (3D) model retrieval method. First, for each 3D model, compute the scale‐invariant heat kernel signature (SI‐HKS) descriptor and the wave kernel signature (WKS) descriptor of each vertex. Then, the normalised weighted bags of phrases feature is obtained and they are fed to the convolutional neural networks. The trust degree of each kind of descriptor is computed, and the total trust degree can be obtained. Finally, the fusion network is trained and the retrieval results can be obtained according to the ranking of the total trust degrees. For the training phase and the testing phase, the authors define different computation methods of the trust degrees and the total trust degrees. The Dempster–Shafer (DS) evidence‐based total trust degrees are used not only in the feature layer but also in the decision layer. The final decision results of the total trust degrees are used in the process of the network learning. So the proposed method can make full use of the complementary information of the SI‐HKS descriptor and the WKS descriptor. Extensive experiments have shown that the proposed multi‐feature fusion method has better performance than a single feature‐based method, and also outperforms other existing state‐of‐the‐art methods.

This paper presents an effective 3D local feature descriptor, which is called Gradient Direction Accumulation-based Heat Kernel Signature (GDA-HKS) descriptor, and its application in nonrigid 3D model retrieval. The GDA-HKS descriptor is based on the heat kernel signature, and it is scale invariant and robust to the nonrigid deformation of the 3D model. Compared with the SI-HKS descriptor, the GDA-HKS descriptor is constructed directly in the time domain, and it can effectively avoid the loss of high frequency information. The absolute gradient difference is used to encode the GDA-HKS descriptor, which can describe the changing trend of the one-dimensional signal more effectively. Extensive experimental results have validated the effectiveness of the designed GDA-HKS descriptor.

Shape and texture information are critical to the accuracy of image classification systems. In this study, the authors propose a novel descriptor called weighted centre‐symmetric local ternary pattern (WCS‐LTP), better characterising the image local texture. Then, based on the proposed WCS‐LTP descriptor, they introduce a new local scale invariant feature transform and WCS‐LTP (SIFT–WCS‐LTP) feature extraction approach. Compared with conventional local CS‐LTP and SIFT features, the authors’ proposed SIFT–WCS‐LTP feature can not only capture the shape information of images, but also tend to extract more precise texture information. Finally, SIFT–WCS‐LTP feature‐based sparse coding spatial pyramid matching (ScSPM) representation classification is proposed for image classification. Extensive experimental results demonstrate that the effectiveness of their proposed SIFT–WCS‐LTP feature‐based ScSPM representation classification algorithm.

This paper proposes a multiscale 3D keypoint detection method based on local weighted dissimilarity measure. At first, compute the local weighted dissimilarity measure of each vertex at different scale. Then determine the detecting scale of each vertex. Finally compare the local weighted dissimilarity measure of each vertex with those of its neighboring surface points at its detecting scale. The keypoint is defined as the vertex that has highest local weighted dissimilarity measure in its neighborhood. The contribution of this paper includes that we propose a novel local weighted dissimilarity measure and the frame of multiscale keypoint detection method. The proposed local weighted dissimilarity measure is computed from the shape index value, and it is invariable to rotation and translation transformation. The multiscale algorithm frame enable the detected key points are robust to noise, especially to high level noise. Extensive experiments have performed to testify the effectiveness of the proposed method.

In this paper, we introduce an improved version of DAISY descriptor algorithm for fast and high-quality image key point matching. Since DAISY descriptor algorithm has many prominent advantages but lacks the ability of handling large in-plane rotation, we thus construct our Rotation-Invariant DAISY descriptor to effectively settle this shortcoming. We first extract key points of images by Harris corner detector, and then use histogram data of every key point to assign a local orientation for each. During the step of constructing descriptors, we rotate the coordinate axis of every key point to its local orientation, and obtain Rotation-Invariant DAISY descriptors. In order to accelerate matching task, we apply PCA to descriptors for dimension reduction. Our experiments demonstrate that the Fast Rotation-Invariant DAISY has better performance in various image distortions. The results also testify that this approach consumes less time when obtains similar results of SIFT descriptor algorithm, and it can be applied more densely in different kinds of images but SIFT cannot.

We present a novel 3D ear reconstruction method based on epipolar geometry. First, we calibrate the camera system. Second, we apply Harris corner detector to extract feature points from the first image. For every feature point, we apply epipolar geometry constrain to draw epipolar line in the second image and search corresponding point on it. Both the feature point and points on the epipolar line are described by DAISY descriptor. By measuring the similarity between the descriptor of the feature point and descriptors of points on the epipolar line using Euclidean distance, we cast away outliers and obtain the matching points. Finally, the 3D ear model can be reconstructed by using triangulation principle. Our method can automatically provide sufficient matching points and it is also less time consuming due to the high performance of DAISY descriptor. Experimental results indicate that our method is feasible.

A novel ear recognition approach is proposed in this paper, which use the SIFT descriptor with global context and the projective invariants to obtain ear features. At first, as the ear images have multiple similar local regions, the SIFT descriptor with global context is used for computing the matching points. This kind of descriptor can discriminate the keypoints with similar local appearances effectively. Then the number of the matching points is used for recognition. Finally, five projective invariants are obtained by computing the cross ratios of five collinear points on the longest axis. Both the number of the matching points and the projective invariants are used for constructing recognition feature. The nearest neighbor method is used for classification. Extensive experiments have performed to valid its efficiency.

This paper presents an automatic 3D ear reconstruction method based on binocular stereo vision. At first, we calibrate the stereo vision system by Zhang's method. Then the quasi-dense matching method is performed. We use SIFT feature based matching approach and the coarse to fine strategy to compute the seed matches. The adapted match propagation algorithm with known epipolar geometry constraint is used for obtain quasi-dense correspondence points. Finally the 3D ear model can be reconstructed by triangulation and the results are optimized using the bundle adjustment technique. Extensive experimental results have shown that our proposed method can obtain denser 3D ear model than Liu's multi-view based method. It can get sufficient 3D ear points with lower cost and higher efficiency.

This paper proposes a novel method based on Haar wavelet transform and uniform local binary patterns (ULBPs) to recognize ear images. Firstly, ear images are decomposed by Haar wavelet transform. Then ULBPs are combined simultaneously with block-based and multi-resolution methods to describe together the texture features of ear sub-images transformed by Haar wavelet. Finally, the texture features are classified by the nearest neighbor method. Experimental results show that Haar wavelet transform can boost effectively up intensity information of texture unit. It is not only fast but also robust to use ULBPs to extract texture features. The recognition rates of the method proposed by this paper outperform remarkably those of the classic PCA or KPCA especially when combining block-based and multi-resolution methods.