Image inpainting attempts to fill in missing areas of corrupted images. Previous works used diverse prior information as constraints to recover high-quality images. Nevertheless, these priors rely on heuristic information and highly empirical selection. Moreover, CNN-based methods ignore the global long-range dependencies between spatial positions in images. This paper presents adaptive prior and long-range dependency-based learners (APLRL) for image inpainting. It mainly constructs an adaptive prior extractor (AdaPE) and an adaptive graph convolution (AdaGConv) operator. Specifically, AdaPE devises a learnable network by integrating partial convolution into residual learning. This enables it to mitigate the pollution of prior information caused by mask influence, effectively learn and extract any unknown explicit and implicit priors in a data-driven manner, and assist in image inpainting. Besides, an AdaGConv operator adaptively learns potential sparse graph structures in images by a learnable threshold strategy, and fuses graph convolution operators to acquire long-distance information on image spatial locations. This improves comprehension of the image’s overall structure and contributes to the network filling in the missing areas more effectively. Experiments reveal the superiority of APLRL over different baselines. Notably, AdaPE provides a readily transferable plug-and-play module. The source code is available at https://github.com/QijinXu/APLRL.

Deep learning has recently gained popularity in hyperspectral unmixing (HU) and typical methods involve convolutional neural network-based (CNN-based) and autoencoding-based methods. However, most existing methods are usually confined to capturing local features in hyperspectral images (HSIs) while neglecting long-range dependency information on spatial position in HSIs, where long-range dependency on spatial positions means the correlations between spatial pixels or regions. Graph neural networks (GNNs) have recently shown great potential in various fields, which model complex spatial relationships and interactions in data. Therefore, this article develops a joint multiscale graph attention and classify-driven autoencoder (MSGA-CD) framework for HU. Its core is to construct a multiscale graph attention abundance (MSGAA) module, a local-global abundance fusion (LGAF) module, and an abundance-classify-driven endmember decoder (ACDE) module. Concretely, MSGAA incorporates a multiscale strategy into the graph attention network (GAT) to extract diverse long-range dependency information on spatial locations in HSIs from different levels and obtain global abundances. Afterward, LGAF integrates local abundance obtained by CNN and global abundance by MSGAA, achieving a more precise abundance representation. Moreover, ACDE clusters all HSI pixel features into various endmember categories using abundance fractions and takes them as priors to drive endmember learning, effectively improving the accuracy of endmember extraction. Finally, the abundance and endmember matrices are trained simultaneously by constraining their dependent relationship through a joint loss. Experiments reveal that MSGA-CD outperforms state-of-the-art methods, offering a promising method for HU.

Point cloud registration is a critical research area in computer vision with extensive applications. Recent studies have unveiled the significant potential of graph neural networks (GNNs) for point cloud registration. One key approach is to leverage the smoothness of graph convolutions to extract similarity information between points. However, as the number of convolution layers increases, the features between points tend to become consistent, and distinctiveness is always neglected, which contradicts point cloud registration. To this end, this paper presents a new GNN framework with 3D graph smoothing-sharpening convolution (GNN-GSSC) for point cloud registration. It includes two new convolutional strategies: graph smoothing convolution (SmoothGConv) and graph sharpening convolution (SharpGConv). The former utilizes Laplacian smoothing to aggregate similar information from neighbouring nodes, whereas the latter encourages each node to move away from its neighbours to obtain more discriminative information. Specifically, we calculate the difference information between the central node and neighbouring nodes to supplement the node feature information while aggregating the similarity information of the nodes. In addition, we devise a Transformer-based overlapping point scoring module, enhancing the emphasis on overlapping areas while weakening the focus on non-overlapping areas by scoring each point. Experiments reveal that the proposed method is optimal compared to other existing methods. More importantly, SharpGConv is a plug-and-play graph convolution module that is particularly advantageous for extracting distinctive information in point cloud registration.

Deep learning-based methods have been widely used in medical image segmentation recently. However, existing works are usually difficult to simultaneously capture global long-range information from images and topological correlations among feature maps. Further, medical images often suffer from blurred target edges. Accordingly, this paper proposes a novel medical image segmentation framework named a label-decoupled network with spatial-channel graph convolution and dual attention enhancement mechanism (LADENet for short). It constructs learnable adjacency matrices and utilizes graph convolutions to effectively capture global long-range information on spatial locations and topological dependencies between different channels in an image. Then a label-decoupled strategy based on distance transformation is introduced to decouple an original segmentation label into a body label and an edge label for supervising the body branch and edge branch. Again, a dual attention enhancement mechanism, designing a body attention block in the body branch and an edge attention block in the edge branch, is built to promote the learning ability of spatial region and boundary features. Besides, a feature interactor is devised to fully consider the information interaction between the body and edge branches to improve segmentation performance. Experiments on benchmark datasets reveal the superiority of LADENet compared to state-of-the-art approaches.

Although convolutional neural networks (CNNs) have shown good performance on grid data, they are limited in the semantic segmentation of irregular point clouds. This article proposes a novel and effective graph CNN framework, referred to as the local-global graph convolutional method (LGGCM), which can achieve short- and long-range dependencies on point clouds. The key to this framework is the design of local spatial attention convolution (LSA-Conv). The design includes two parts: generating a weighted adjacency matrix of the local graph composed of neighborhood points, and updating and aggregating the features of nodes to obtain the spatial geometric features of the local point cloud. In addition, a smooth module for central points is incorporated into the process of LSA-Conv to enhance the robustness of the convolution against noise interference by adjusting the position coordinates of the points adaptively. The learned robust LSA-Conv features are then fed into a global spatial attention module with the gated unit to extract long-range contextual information and dynamically adjust the weights of features from different stages. The proposed framework, consisting of both encoding and decoding branches, is an end-to-end trainable network for semantic segmentation of 3-D point clouds. The theoretical analysis of the approximation capabilities of LSA-Conv is discussed to determine whether the features of the point cloud can be accurately represented. Experimental results on challenging benchmarks of the 3-D point cloud demonstrate that the proposed framework achieves excellent performance.

Graph neural networks (GNNs) have a powerful ability to capture long-range spatial correlations in hyperspectral images (HSIs). However, existing GNN-based HSI classification methods are vulnerable to hand-crafted graphs, as the manner in which these graphs are constructed are often inappropriate and are likely to violate intrinsic graph properties, such as sparsity and low-rank. More importantly, the goal of HSI classification is to categorize each individual pixel into a land-cover class, but existing methods usually overuse global dependencies and ignore the importance of individualized spectral characteristics. Therefore, this paper proposes a Global-to-Local decoupling Graph network (GoLoG) to conduct HSI classification in a global-to-local framework, which jointly optimizes the graph structure and network parameters guided by both intrinsic graph properties and classification loss. Specifically, a novel global-to-local network framework with successive global and local graph convolutional stages is constructed. By decoupling global and local stages, global contextual information can be exploited, and the individualized information of each hyperspectral pixel can be emphasized for HSI classification. Second, a sparse and low-rank graph structure learning model is proposed to refine and renovate the initial-construct graph. Finally, to unify graph structure learning and network training, a joint alternating update algorithm is introduced to jointly optimize the sparse and low-rank graph structure learning model and the global-to-local network framework. Extensive experiments demonstrate that the proposed GoLoG has obvious advantages compared with other state-of-the-art HSI classification methods.

The use of deep learning methods in hyperspectral image (HSI) classification has been a promising approach due to its powerful ability to automatically extract features in recent years. This article proposes a novel deep framework for HSI classification problems, referred to as feature-grouped network based on spectral–spatial connected attention mechanism (FG-SSCA). Different from the existing deep learning methods, the proposed framework integrates the spectral attention module and spatial attention module continuously from the raw HSI input, which is embedded into convolutional neural networks and could enhance the distinguishing ability of spectral bands and learn the spatial relevance between the neighboring pixels together. Meanwhile, the generating feature maps are sliced into a series of small groups in sequence along the direction of spectral bands and each group sequentially extracts spatial–spectral features through multiple spectral and spatial residual blocks. This feature-grouped strategy could fully utilize the redundancy and difference of bands and obtain more available and valuable information. The proposed FG-SSCA method could greatly improve generalization performance and make tremendous successes in HSI classification. Experimental results on several HSI benchmark data sets verify the effectiveness and superiority of the proposed method in comparison with the state-of-the-art approaches for HSI classification.

Hyperspectral image (HSI) classification has always been one of the hot issues in the study of geographic remote sensing information, and graph neural networks have attracted much attention in recent years. Several graph neural network-based approaches have been introduced into HSI study to explore the spatial information of HSI within a constructed graph. However, the existing methods of building HSI-based graphs are always unsuitable and inaccurate due to the complicated spatial variability of spectral signatures. Meanwhile, these graph-based HSI classification methods usually suffer from the over-smoothing problem. Motivated by these, this article presents a novel deep network with adaptive graph structure integration (DNAGSI), which could learn a graph structure of HSI dynamically and promote its discriminative ability with devising a much deeper network architecture. Specifically, dynamic graphs are first built across different layers and adaptively integrated with the initial graph structure to boost the robust graph representation of HSI. Second, the initial residual and identity mapping are employed to significantly increase the depth of the network and obtain more abstract deep features. Finally, a joint loss with center loss is devised to learn the similarity relationship between hyperspectral pixels explicitly, thereby gathering the intraclass graph features. Benefiting from the integration of center loss, initial residual, and identity mapping, the proposed method can alleviate the over-smoothing problem effectively to some extent. Experiments on benchmark HSI datasets demonstrate the superiority of DNAGSI over state-of-the-art methods.

In recent years, as an important means of digital image copyright protection, image invisible watermarking technology has attracted more and more attention. Based on the watermarking system built by deep learning model, this paper proposes an optimization algorithm which can extract watermark and improve robustness. The watermark is embedded and extracted by the encoder and decoder of the model respectively. In order to make the model robust to various image attacks, two noise layers are added to the model. In order to further improve the robustness and concealment of watermark, adversarial training is used. In order to improve the accuracy of model joint training, a two-step training strategy is adopted. The first step is to train the best pre training model of encoder and decoder. The second step is to train different decoders according to different image attack means, so as to improve the robustness of the whole watermarking system. Experimental results show that, compared with other watermarking algorithms based on deep learning model, the proposed method is superior to other algorithms in the concealment and robustness.

This article proposes a novel regularization deep cascade broad learning system (DCBLS) architecture, which includes one cascaded feature mapping nodes layer and one cascaded enhancement nodes layer. Then, the transformation feature representation is easily obtained by incorporating the enhancement nodes and the feature mapping nodes. Once such a representation is established, a final output layer is constructed by implementing a simple convex optimization model. Furthermore, a parallelization framework on the new method is designed to make it compatible with large-scale data. Simultaneously, an adaptive regularization parameter criterion is adopted under some conditions. Moreover, the stability and error estimate of this method are discussed and proved mathematically. The proposed method could extract sufficient available information from the raw data compared with the standard broad learning system and could achieve compellent successes in image denoising. The experiments results on benchmark datasets, including natural images as well as hyperspectral images, verify the effectiveness and superiority of the proposed method in comparison with the state-of-the-art approaches for image denoising.

Mixture noise removal is a fundamental problem in hyperspectral images' (HSIs) processing that holds significant practical importance for subsequent applications. This problem can be recast as an approximation issue of a low-rank matrix. In this paper, a novel smooth rank approximation (SRA) model is proposed to cope with these mixture noises for HSIs. The crux idea is to devise a general smooth function under some assumptions to directly approximate the rank function, which attempts to explore a closer approximation than conventional methods. This new optimization model can be easily solved by the convex analysis tool and can remove the mixture noises of HSIs quickly and effectively. Subsequently, we give a feasible iterative algorithm, and the corresponding convergence analysis is discussed mathematically. Experimental results from the simulated data set as well as real data sets illustrate that the proposed SRA method significantly outperforms the state-of-the-art methods on HSI denoising.

Matrix completion has been widely used in image processing, in which the popular approach is to formulate this issue as a general low-rank matrix approximation problem. This paper proposes a novel regularization method referred to as truncated Frobenius norm (TFN), and presents a hybrid truncated norm (HTN) model combining the truncated nuclear norm and truncated Frobenius norm for solving matrix completion problems. To address this model, a simple and effective two-step iteration algorithm is designed. Further, an adaptive way to change the penalty parameter is introduced to reduce the computational cost. Also, the convergence of the proposed method is discussed and proved mathematically. The proposed approach could not only effectively improve the recovery performance but also greatly promote the stability of the model. Meanwhile, the use of this new method could eliminate large variations that exist when estimating complex models, and achieve competitive successes in matrix completion. Experimental results on the synthetic data, real-world images as well as recommendation systems, particularly the use of the statistical analysis strategy, verify the effectiveness and superiority of the proposed method, i.e. the proposed method is more stable and effective than other state-of-the-art approaches.