Recent advances in the design of convolutional neural networks have shown that performance can be enhanced by improving the ability to represent multi-scale features. However, most existing methods either focus on designing more sophisticated attention modules, which leads to higher computational costs, or fail to effectively establish long-range channel dependencies, or neglect the extraction and utilization of structural information. This work introduces a novel module, the Multi-Branch Concatenation (MBC), designed to process input tensors and extract multi-scale feature maps. The MBC module introduces new degrees of freedom (DoF) in the design of attention networks by allowing for flexible adjustments to the types of transformation operators and the number of branches. This study considers two key transformation operators: multiplexing and splitting, both of which facilitate a more granular representation of multi-scale features and enhance the receptive field range. By integrating the MBC with an attention module, a Multi-Branch Attention (MBA) module is developed to capture channel-wise interactions within feature maps, thereby establishing long-range channel dependencies. Replacing the 3x3 convolutions in the bottleneck blocks of ResNet with the proposed MBA yields a new block, the Efficient Multi-Branch Attention (EMBA), which can be seamlessly integrated into state-of-the-art backbone CNN models. Furthermore, a new backbone network, named EMBANet, is constructed by stacking EMBA blocks. The proposed EMBANet has been thoroughly evaluated across various computer vision tasks, including classification, detection, and segmentation, consistently demonstrating superior performance compared to popular backbones.

Phishing URL detection has become a critical challenge in cybersecurity, with existing methods often struggling to maintain high accuracy while generalizing across diverse datasets. In this article, we introduce BERT-PhishFinder, a novel and efficient transformer-based model designed to tackle this problem. While most traditional approaches rely heavily on lexical features or complex convolutional architectures, BERT-PhishFinder leverages the power of DistilBERT, a lightweight yet highly effective transformer, to capture rich contextual representations of URL sequences. To enhance the model’s robustness and reduce overfitting, we strategically incorporate SpatialDropout1D in the embedding layers, along with global average pooling and global max pooling techniques to extract both comprehensive and key discriminative features. The pooled representations are thoughtfully concatenated to form a comprehensive feature representation. Through this carefully crafted design, our model adopts ensemble learning, as it undergoes multiple parallel dense layers, each with distinct parameters and dropout regularization. This facilitates learning diverse patterns and features from the input URL sequence, culminating in exceptional phishing URL detection performance. Extensive evaluations against conventional deep learning algorithms, transformer models (XLNet, RoBERTa, ALBERT), and other existing methods on five benchmark datasets show that BERT-PhishFinder not only achieves the state of-the-art real phishing URL detection but also accomplishes this with reduced label dependency.

Generalizable semantic segmentation, aims to excel on unseen target domains, as a critical focus due to the widespread practical applications requiring high generalizability. Class-wise prototypes, which depict class-wise centroids, as a type of domain-invariant information are key to improving the model generalizability due to its stability and representativeness. However, this manner faces some challenges. First, the existing methods adopt a coarse prototypical alignment form, potentially compromising performance. Second, the naive prototype generally serves as the class centroid generated by an average operation from source data batches, risks source domain overfitting, and may be detrimentally impacted by unrelated source data. Third, from a broader perspective, rather than just from a prototypical alignment perspective, the existing methods treat all samples equally, which is against the conclusion that different source features have different adaptation difficulties. To tackle these issues, we propose a novel method for generalizable semantic segmentation called Prototypical Progressive Alignment and Reweighting (PPAR) depending on the strong generalized representation of the Contrastive Language-Image Pretraining (CLIP) model. In particular, we first define the Original Text Prototype (OTP) and Visual Text Prototype (VTP) generated by the CLIP model, laying the foundation for the subsequent effective alignment strategy. Then, we propose a prototypical progressive alignment strategy by an easy-to-difficult alignment form to reduce domain-variant information progressively instead of directly. Finally, we propose a prototypical reweighting learning strategy that estimates the importance of the source data and corrects its learning weight to alleviate the influence of unrelated source features, i.e. alleviate negative transfer. Moreover, we also offer a theoretical insight into our method and it shows that our method compiles well on the domain generalization theory. Extensive experiments on several popular datasets demonstrate that our PPAR method achieves superior performance, proving the effectiveness of our method. The source code will available at: https://github.com/Hectoor/PPAR

Unsupervised domain adaptative semantic segmentation is a powerful solution for the distribution shift problem between the source and target domains. However, such methods need specified target domain data that may be unavailable in actual applications due to excess expensive collection. Generalizable semantic segmentation as a new paradigm appears in recent research, which aims to generalize well on distinct unseen domains only using source domain data. The existing methods focus on learning domain-invariant features by using global distribution alignment strategies, which may lead to a decreased discriminability of the model. To cope with this challenge, we propose a fine-grained self-supervision (FGSS) framework for generalizable semantic segmentation that takes into account both discriminability and generalizability from the perspective of the intra-class relationship. The FGSS framework contains single-view and multi-view versions. In the single-view version, we propose a fine-grained self-supervision strategy to distinguish the sub-parts of the semantic class for better class discriminability. In the multi-view version, we propose a class prototype feature enhancement strategy to generate another view (i.e. another representation of the original representation). Then, we propose a multi-view mutual supervision loss to enforce consistency between different views and further enhance the generalizability of the model. Experimental results on five widely-used datasets, i.e., GTAV, SYNTHIA, BDD100K, Cityscapes, and Mapillary, demonstrate that our FGSS framework achieves superior performance compared to state-of-the-art methods.

Current deep models excel in object detection for classification and localization. However, precise object rotation estimation within the visual context of an input image remains underexplored due to the lack of object datasets with rotation annotations. This paper addresses these challenges by tackling rotation estimation for parked bikes with respect to their parking area. Firstly, 3D graphics were leveraged to build a camera-agnostic well-annotated Synthetic Bike Rotation Dataset (SynthBRSet). Subsequently, an object-to-spot rotation estimator (OSRE) is introduced by extending object detection to regress bike rotations in two axes. As the proposed model trained purely on synthetic data, image smoothing techniques adopted during deployment on real-world images. The proposed OSRE has undergone evaluation on both synthetic and real-world data, showing promising results. Our data and code are available at https://saghiralfasly.github.io/OSRE-Project/.

Convolutional neural networks (CNNs) have come to dominate vision-based deep neural network structures in both image and video models over the past decade. However, convolution-free vision Transformers (ViTs) have recently outperformed CNN-based models in image recognition. Despite this progress, building and designing video Transformers have not yet obtained the same attention in research as image-based Transformers. While there have been attempts to build video Transformers by adapting image-based Transformers for video understanding, these Transformers still lack efficiency due to the large gap between CNN-based models and Transformers regarding the number of parameters and the training settings. In this work, we propose three techniques to improve video understanding with video Transformers. First, to derive better spatiotemporal feature representation, we propose a new spatiotemporal attention scheme, termed synchronized spatiotemporal and spatial attention (SSTSA), which derives the spatiotemporal features with temporal and spatial multiheaded self-attention (MSA) modules. It also preserves the best spatial attention by another spatial self-attention module in parallel, thereby resulting in an effective Transformer encoder. Second, a motion spotlighting module is proposed to embed the short-term motion of the consecutive input frames to the regular RGB input, which is then processed with a single-stream video Transformer. Third, a simple intraclass frame interlacing method of the input clips is proposed that serves as an effective video augmentation method. Finally, our proposed techniques have been evaluated and validated with a set of extensive experiments in this study. Our video Transformer outperforms its previous counterparts on two well-known datasets, Kinetics400 and Something-Something-v2.

In this article, a distributed optimization algorithm is proposed for solving a distributed resource allocation problem (DRAP) with general inequality and heterogeneous coupled equality constraints. The communication topologies herein are considered to be jointly connected and directed interacted. To deal with the effects of inequality constraints, an adaptive item of updating penalty gain on-line is introduced in algorithmic design, which enforces the state enter to the constraint sets dynamically. Further, with the aid of Lyapunov method, the convergence to global optimal solution of nonsmooth DRAP is obtained. To effectively alleviate the communication burden caused by frequent interactions, an event-triggered mechanism is proposed to drive the agents with free-initial state, while also ensuring the exclusion of Zeno behavior. Compared with existing algorithms for DRAP, the time-varying auxiliary function designed in distributed algorithms herein avoids the preemptive estimation of global parameters that may cause the failure of the distributed framework, including the global Lipschitz coefficients of objective functions and the eigenvalues of full Laplacian matrix. Finally, numerical simulations and application of economic dispatch in smart grid illustrate the validity of designed algorithm.

In the study of salient object detection, multiview features play an important role in identifying various underlying salient objects. As to current common patch-based methods, all different features are handled directly by stacking them into a high-dimensional vector to represent related image patches. These approaches ignore the correlations inhering in the original spatial structure, which may lead to the loss of certain underlying characterization such as view interaction. In this article, different from currently available approaches, a tensorial feature representation framework is developed for the salient object detection in order to better explore the complementary information of multiview features. Under the tensor framework, a tensor low-rank constraint is applied to the background to capture its intrinsic structure, a tensor group sparsity regularization is posed on the salient part, and a tensorial sliced Laplacian regularization is then introduced to enlarge the gap between the subspaces of the background and salient object. Moreover, a nonconvex tensor Log-determinant function, instead of the tensor nuclear norm, is adopted to approximate the tensor rank for effectively suppressing the confusing information resulted from underlying complex backgrounds. Further, we have deduced the closed-form solution of this nonconvex minimization problem and established a feasible algorithm whose convergence is mathematically proven. Experiments on five well-known public datasets are provided and the simulations demonstrate that our method outperforms the latest unsupervised handcrafted features-based methods in the literature. Furthermore, our model is flexible with various deep features and is competitive with the state-of-the-art approaches.

Semantic segmentation assigns a category for each pixel and has achieved great success in a supervised manner. However, it fails to generalize well in new domains due to the domain gap. Domain adaptation is a popular way to solve this issue, but it needs target data and cannot handle unavailable domains. In domain generalization (DG), the model is trained without the target data and DG aims to generalize well in new unavailable domains. Recent works reveal that shape recognition is beneficial for generalization but still lack exploration in semantic segmentation. Meanwhile, the object shapes also exist a discrepancy in different domains, which is often ignored by the existing works. Thus, we propose a Shape-Invariant Learning (SIL) framework to focus on learning shape-invariant representation for better generalization. Specifically, we first define the structural edge, which considers both the object boundary and the inner structure of the object to provide more discrimination cues. Then, a shape perception learning strategy including a texture feature discrepancy reduction loss and a structural feature discrepancy enlargement loss is proposed to enhance the shape perception ability of the model by embedding the structural edge as a shape prior. Finally, we use shape deformation augmentation to generate samples with the same content and different shapes. Essentially, our SIL framework performs implicit shape distribution alignment at the domain-level to learn shape-invariant representation. Extensive experiments show that our SIL framework achieves state-of-the-art performance.

With the assumption that a video dataset is multimodality annotated in which auditory and visual modalities both are labeled or class-relevant, current multimodal methods apply modality fusion or cross-modality attention. However, effectively leveraging the audio modality in vision-specific annotated videos for action recognition is of particular challenge. To tackle this challenge, we propose a novel audio-visual framework that effectively leverages the audio modality in any solely vision-specific annotated dataset. We adopt the language models (e.g., BERT) to build a semantic audio-video label dictionary (SAVLD) that maps each video label to its most K-relevant audio labels in which SAVLD serves as a bridge between audio and video datasets. Then, SAVLD along with a pretrained audio multi-label model are used to estimate the audio-visual modality relevance during the training phase. Accordingly, a novel learnable irrelevant modality dropout (IMD) is proposed to completely drop out the irrelevant audio modality and fuse only the relevant modalities. Moreover, we present a new two-stream video Transformer for efficiently modeling the visual modalities. Results on several vision-specific annotated datasets including Kinetics400 and UCF-101 validated our framework as it outperforms most relevant action recognition methods.

Recently, blur detection is a hot topic in computer vision. It can accurately segment the blurred areas from an image, which is conducive for the post-processing of the image. Although many hand-crafted features based approaches have been presented during the last decades, they were not robust to the complex scenarios. To solve this problem, we newly establish a boundary-aware multi-scale deep network in this paper. First, the VGG-16 network is used to extract the deep features from multi-scale layers. Contrast layers and deconvolutional layers are added to make the difference between the blurred areas and clear areas more prominent. At last, a new boundary-aware penalty is introduced, which makes the edges of our results much clearer. Our method spends about 0.2 s to evaluate an image. Experiments on the large dataset confirm that the proposed model performs better than other models.

Defocus blur detection aims at separating regions on focus from out-of-focus for image processing. With today’s popularity of mobile phones with portrait mode, accurate defocus blur detection has received more and more attention. There are many challenges that we currently confront, such as blur boundaries of defocus regions, interference of messy backgrounds and identification of large flat regions. To address these issues, in this paper, we propose a new deep neural network with both global and local pathways for defocus blur detection. In global pathway, we locate the objects on focus by semantical search. In local pathway, we refine the predicted blur regions via multi-scale supervisions. In addition, the refined results in local pathway are fused with searching results in global pathway by a simple concatenation operation. The structure of our new network is developed in a feasible way and its function appears to be quite effective and efficient, which is suitable for the deployment on mobile devices. It takes about 0.2[Formula: see text]s per image on a regular personal laptop. Experiments on both CUHK dataset and our newly proposed Defocus400 dataset show that our model outperforms existing state-of-the-art methods.

The strong correlation between neurons or filters can significantly weaken the generalization ability of neural networks. Inspired by the well-known Tammes problem, we propose a novel diversity regularization method to address this issue, which makes the normalized weight vectors of neurons or filters distributed on a hypersphere as uniformly as possible, through maximizing the minimal pairwise angles (MMA). This method can easily exert its effect by plugging the MMA regularization term into the loss function with negligible computational overhead. The MMA regularization is simple, efficient, and effective. Therefore, it can be used as a basic regularization method in neural network training. Extensive experiments demonstrate that MMA regularization is able to enhance the generalization ability of various modern models and achieves considerable performance improvements on CIFAR100 and TinyImageNet datasets. In addition, experiments on face verification show that MMA regularization is also effective for feature learning. Code is available at: https://github.com/wznpub/MMA_Regularization.

Recently, induced by incorporating the ubiquitous data collected by AI application, there are more and more demand for transferring data to cloud servers due to the restriction of computing resources of devices. To transfer image data efficiently, neural network (NN) codec can be a wise choice, which can result in higher ratio of compression with similar image quality compared with conventional codec methods. However, when the NN codec is employed in the communication system, it is likely that it can be disturbed by channel distortions. In this paper, we innovatively propose a norm method to measure the NN codec's robustness for certain tasks. And with the aid of this method, we develop a greedy algorithm using gradients to find adversarial samples of certain tasks considering communication system distortions, i.e. channel distortions during compressed image or video transferring to cloud server systems. Aided by the adversarial samples, the NN codec has been proved to have better tolerance of communication distortions that the Top-1 accuracy of the image classification can be improved 1.8%.

The goal of this paper is to develop a novel numerical method for efficient multiplicative noise removal. The nonlocal self-similarity of natural images implies that the matrices formed by their nonlocal similar patches are low-rank. By exploiting this low-rank prior with application to multiplicative noise removal, we propose a nonlocal low-rank model for this task and develop a proximal alternating reweighted minimization (PARM) algorithm to solve the optimization problem resulting from the model. Specifically, we utilize a generalized nonconvex surrogate of the rank function to regularize the patch matrices and develop a new nonlocal low-rank model, which is a nonconvex nonsmooth optimization problem having a patchwise data fidelity and a generalized nonlocal low-rank regularization term. To solve this optimization problem, we propose the PARM algorithm, which has a proximal alternating scheme with a reweighted approximation of its subproblem. A theoretical analysis of the proposed PARM algorithm is conducted to guarantee its global convergence to a critical point. Numerical experiments demonstrate that the proposed method for multiplicative noise removal significantly outperforms existing methods, such as the benchmark SAR-BM3D method, in terms of the visual quality of the denoised images, and of the peak-signal-to-noise ratio (PSNR) and the structural similarity index measure (SSIM) values.

Most of the discriminative feature learning methods are specifically developed for metric learning, however, the effectiveness may be not obvious for other tasks. In this letter, we propose a novel discrimination regularization method for image classification, which enhances the intra-class compactness and inter-class discrepancy simultaneously, through decreasing the minimalangle (DMA) between the feature vector and any one of the weight vectors in classification layer. This method can robustly improve the discriminability and generalizability of neural networks and easily exert its effect by plugging the DMA regularization term into the loss function with negligible computational overhead. The DMA regularization is simple, efficient, and effective. Therefore, it can be used as a basic regularization method for models based on neural networks. We evaluate DMA by applying it to various modern models on CIFAR10, CIFAR100, and TinyImageNet datasets, decreasing the test error rate by 0.2-0.4%, 0.2-1.5%, and 0.3-0.4% respectively. Code is available at: https://github.com/wznpub/DMA_Regularization.

Capsule projection network (CapProNet) has shown its ability to obtain semantic information, and spatial structural information from the raw images. However, the vector capsule of CapProNet has limitations in representing semantic information due to ignoring local information. Besides, the number of trainable parameters also increases greatly with the dimension of the feature vector. To that end, we propose a matrix capsule convolution projection (MCCP) module by replacing the feature vector with a feature matrix, of which each column represents a local feature. The feature matrix is then convoluted by columns into capsule subspaces to decrease the number of trainable parameters effectively. Furthermore, the CapDetNet is designed to explore the structural information encoding of the MCCP module based on object detection task. Experimental results demonstrate that the proposed MCCP outperforms the baselines in image classification, and CapDetNet achieves the 2.3% performance gain in object detection.

This paper presents the multistability analysis of almost periodic state solutions for memristive Cohen-Grossberg neural networks (MCGNNs) with both distributed delay and discrete delay. The activation function of the considered MCGNNs is generalized to be nonmonotonic and nonpiecewise linear. It is shown that the MCGNNs with n-neuron have (K + 1)nlocally exponentially stable almost periodic solutions, where nature number K depends on the geometrical structure of the considered activation function. Compared with the previous related works, the number of almost periodic state solutions of the MCGNNs is extensively increased. The obtained conclusions in this paper are also capable of studying the multistability of equilibrium points or periodic solutions of the MCGNNs. Moreover, the enlarged attraction basins of attractors are estimated based on original partition. Some comparisons and convincing numerical examples are provided to substantiate the superiority and efficiency of obtained results.

The emergence of Internet of Things (IoT) and new manufacturing paradigms have brought greater complexity of massive datasets. Radio frequency identification (RFID), as one of the key IoT technologies, has been used to collect real-time production data to support the manufacturing decision-making in smart factories. The adoption of these technologies results in a large amount of data collection. To extract useful information from this data, this paper utilizes a big data approach to figure out useful insights from RFID-enabled data regarding possible bottlenecks or inefficiencies on the shop floor so as to improve the quality management. Time and quality are the main metrics measured in this paper, where the longest process times, part accuracy percentage, and failure rate are determined for each of the workers (UserIDs) and process types (ProcCodes). Key findings and observations are significant to make advanced decisions in the smart factory by making full use of the RFID captured data.

In this paper, we analyze the inner product of weight vector w and data vector x in neural networks from the perspective of vector orthogonal decomposition and prove that the direction gradient of w decreases with the angle between them close to 0 or π. We propose the Projection and Rejection Product (PR Product) to make the direction gradient of w independent of the angle and consistently larger than the one in standard inner product while keeping the forward propagation identical. As a reliable substitute for standard inner product, the PR Product can be applied into many existing deep learning modules, so we develop the PR Product version of fully connected layer, convolutional layer and LSTM layer. In static image classification, the experiments on CIFAR10 and CIFAR100 datasets demonstrate that the PR Product can robustly enhance the ability of various state-of-the-art classification networks. On the task of image captioning, even without any bells and whistles, our PR Product version of captioning model can compete or outperform the state-of-the-art models on MS COCO dataset. Code has been made available at: https://github.com/wzn0828/PR_Product.

In the fusion process of medical computed tomography (CT) and magnetic resonance image (MRI), traditional multiscale methods often reduce the contrast of fused images. Although sparse representation (SR) methods overcome this shortcoming, they are often too smooth along the strong edges of the fusion image. To overcome these shortcomings, CT and MRI image fusion based on multiscale decomposition method and hybrid approach is proposed. There are three main steps. First, the cartoon parts and texture parts of CT and MRI are obtained by the improved image decomposition method using global sparse gradients. Second, the large structure cartoon parts are fused using the specific cartoon dictionary and the ‘L1‐max norm’ principle. The textured parts are fused using non‐subsampled contourlet transformation (NSCT) and the maximum energy rule. Finally, the final result is obtained by superimposing the fused cartoon part and the fused texture part. The experimental results demonstrate that the proposed method outperforms the state‐of‐the‐art method SR and NSCT in terms of visual effect and objective quality.

Clustering aims at naturally grouping the data according to the underlying data distribution. The data distribution is often estimated using a parametric or nonparametric model, e.g., Gaussian mixture or kernel density estimation. Compared with nonparametric models, parametric models are statistically stable, i.e., a small perturbation of data points leads to a small change in the estimated density. However, parametric models are highly sensitive to outliers because the data distribution is far away from the parametric assumptions in the presence of outliers. Given a parametric clustering algorithm, this paper shows how to turn this algorithm into a robust one. The idea is to modify the original parametric density into a semiparametric one. The high-density data that form the core of each cluster are modeled with the original parametric density. The low-density data are often far away from the cluster cores and may have an arbitrary shape, thus are modeled using a nonparametric density. A combination of parametric and nonparametric clustering algorithms is used to group the data modeled as a semiparametric density. From the robust statistical point of view, the proposed method has good robustness properties. We test the proposed algorithm on several synthetic and 70 UCI data sets. The results indicate that the semiparametric method could significantly improve the clustering performance.

Non-negative Matrix Factorization (NMF), as a promising image-data representation approach, encounters the problems of slow convergence and weak classification ability. To overcome these limitations, this paper, based on different error measurements, proposes two kinds of NMF algorithms with fast gradient descent and high discriminant performance. It is shown that the proposed Fast NMF (FNMF) methods have larger step sizes than those of traditional NMFs. Moreover, the traditional NMFs are the special cases of our methods. To further enhance the discriminative power of non-negative features, we exploit our previous block NMF technique and obtain Block FNMF (BFNMF) algorithms, which are supervised decomposition approaches with some good properties, such as the highly sparse features and orthogonal features from different classes. In experiments, both convergence on non-negative decomposition and performance on face recognition (FR) are considered for evaluations. Compared with traditional NMF algorithms and some state-of-the-art methods, experimental results indicate the effective and superior performance of the proposed NMF methods.

Capsule network is a novel architecture to encode the properties and spatial relationships of the feature in an image, which shows encouraging results on image classification. However, the original capsule network is not suitable for some classification tasks, where the target objects are complex internal representations. Hence, we propose a multi-scale capsule network that is more robust and efficient for feature representation in image classification. The proposed multi-scale capsule network consists of two stages. In the first stage, structural and semantic information are obtained by multi-scale feature extraction. In the second stage, the hierarchy of features is encoded to multi-dimensional primary capsules. Moreover, we propose an improved dropout to enhance the robustness of the capsule network. Experimental results show that our method has a competitive performance on FashionMNIST and CIFAR10 datasets.

This technical correspondence considers finite-time synchronization of dynamical networks by designing aperiodically intermittent pinning controllers with logarithmic quantization. The control scheme can greatly reduce control cost and save both communication channels and bandwidth. By using multiple Lyapunov functions and convex combination techniques, sufficient conditions formulated by a set of linear matrix inequalities are derived to guarantee that all the node systems are synchronized with an isolated trajectory in a finite settling time. Compared with existing results, the main characteristics of this paper are twofold: 1) quantized controller is used for finite-time synchronization and 2) the designed multiple Lyapunov functions are strictly decreasing. An optimal algorithm is proposed for the estimation of settling time. Numerical simulations are provided to demonstrate the effectiveness of the theoretical analysis.

The generalization ability of a classifier learned from a training set is usually dependent on the classifier's uncertainty, which is often described by the fuzziness of the classifier's outputs on the training set. Since the exact dependency relation between generalization and uncertainty of a classifier is quite complicated, it is difficult to clearly or explicitly express this relation in general. This paper shows a specific study on this relation from the viewpoint of complexity of classification by choosing extreme learning machines as the classification algorithms. It concludes that the generalization ability of a classifier is statistically becoming better with the increase of uncertainty when the complexity of the classification problem is relatively high, and the generalization ability is statistically becoming worse with the increase of uncertainty when the complexity is relatively low. This paper tries to provide some useful guidelines for improving the generalization ability of classifiers by adjusting uncertainty based on the problem complexity.

This paper presents a novel model for video salient object detection called spatiotemporal constrained optimization model (SCOM), which exploits spatial and temporal cues, as well as a local constraint, to achieve a global saliency optimization. For a robust motion estimation of salient objects, we propose a novel approach to modeling the motion cues from optical flow field, the saliency map of the prior video frame and the motion history of change detection, which is able to distinguish the moving salient objects from diverse changing background regions. Furthermore, an effective objectness measure is proposed with intuitive geometrical interpretation to extract some reliable object and background regions, which provided as the basis to define the foreground potential, background potential, and the constraint to support saliency propagation. These potentials and the constraint are formulated into the proposed SCOM framework to generate an optimal saliency map for each frame in a video. The proposed model is extensively evaluated on the widely used challenging benchmark data sets. Experiments demonstrate that our proposed SCOM substantially outperforms the state-of-the-art saliency models.

In this paper, based on calculus and penalty method, a one-layer recurrent neural network is proposed for solving constrained complex-variable convex optimization. It is proved that for any initial point from a given domain, the state of the proposed neural network reaches the feasible region in finite time and converges to an optimal solution of the constrained complex-variable convex optimization finally. In contrast to existing neural networks for complex-variable convex optimization, the proposed neural network has a lower model complexity and better convergence. Some numerical examples and application are presented to substantiate the effectiveness of the proposed neural network.

This paper investigates the problem of finite-time drive-response synchronization for a class of neural networks with discontinuous activations, time-varying discrete and infinite-time distributed delays, and mismatched parameters. In order to cope with the difficulties induced by discontinuous activations, time delays, as well as mismatched parameters simultaneously, new 1-norm-based analytical techniques are developed. Both state feedback and adaptive controllers with and without the sign function are designed. Based on differential inclusion theory and Lyapunov functional method, several sufficient conditions on the finite-time synchronization are obtained. Our results show that the controllers with a sign function can reduce the conservativeness of control gains and the controllers without a sign function can overcome the chattering phenomenon. Numerical examples are given to show the effectiveness of the theoretical analysis.

In the past five years, salient object detection has become one of the hot topics in the field of computer vision. Focus is a naturally strong indicator for the salient object detection task, but is not well studied. In this paper, a novel method is proposed to estimate the focus prior map for an arbitrary image. Different from the current edge density estimation based methods, the proposed method is based on the sparse defocus dictionary learning on a newly designed dataset. The focus strength is measured by the number of non-zero coefficients of the dictionary atoms. Objectness proposal method is introduced to improve the performance. Comparison with the other focusness estimation methods, the proposed focus prior map is more accurate and easier to be integrated by the other salient object detection methods. Experiments have confirmed the effectiveness and importance of the proposed focus prior.

Image denoising as a part of pre-processing in image analysis is a challenging area of research since noise removal and image detail preservation need a tradeoff. For classical denoising models, the convex total variation (TV) or some nonconvex regularizers are used to achieve the tradeoff. However, the denoising performance of classical models is still inadequate. To overcome this problem, this paper proposes a new variational model for image restoration, where a weighted regularizer is designed to protect more geometric structural details of images from over-smoothing and to remove much noise simultaneously. To solve the model efficiently, a novel algorithm based on Chambolle’s dual projection method and the iteratively reweighting method is presented. Numerical results prove that the proposed denoising method can show better performance than the classical TV-based and the nonconvex regularizer-based denoising methods.

Multiple-instance active learning (MIAL) is a paradigm to collect sufficient training bags for a multiple-instance learning (MIL) problem, by selecting and querying the most valuable unlabeled bags iteratively. Existing works on MIAL evaluate an unlabeled bag by its informativeness with regard to the current classifier, but neglect the internal distribution of its instances, which can reflect the diversity of the bag. In this paper, two diversity criteria, i.e., clustering-based diversity and fuzzy rough set based diversity, are proposed for MIAL by utilizing a support vector machine (SVM) based MIL classifier. In the first criterion, a kernel k-means clustering algorithm is used to explore the hidden structure of the instances in the feature space of the SVM, and the diversity degree of an unlabeled bag is measured by the number of unique clusters covered by the bag. In the second criterion, the lower approximations in fuzzy rough sets are used to define a new concept named dissimilarity degree, which depicts the uniqueness of an instance so as to measure the diversity degree of a bag. By incorporating the proposed diversity criteria with existing informativeness measurements, new MIAL algorithms are developed, which can select bags with both high informativeness and diversity. Experimental comparisons demonstrate the feasibility and effectiveness of the proposed methods.

A color design process for fabric images can resort to a solution of a color transfer problem based on given color themes. Usually, the color transfer process contains an image segmentation phase and an image construction phase. In this paper, a novel color transfer method for fabric images is proposed. Compared with classical color transfer methods, the new method has the following three main innovations. First, the new method, in its image segmentation phase, follows an assumption that a fabric image can be decomposed into cartoon and texture components, which means the new color transfer method, in its image segmentation, phase incorporates an image decomposition process. The advantage of the innovation is that the cartoon component is more suitable than the original image to be used to partition the fabric image. Second, the new color transfer method can generate more vivid color transfer results since the above texture component is used to describe yarn texture details in the image construction phase. Third, the total generalized variation (TGV) regularizer is used to further improve the performance of image decomposition. Here, the TGV regularizer is good at estimating the weak lightness variation of the cartoon component with the CIELab color scheme. In addition, by using the augmented Lagrange multiplier method, we derive an efficient algorithm to search for the solutions to the proposed color transfer problem. Numerical results demonstrate that the proposed color transfer method can generate better results for fabric images.

Choosing suitable kernels plays an important role in the performance of kernel methods. Recently, a number of studies were devoted to developing nonparametric kernels. Without assuming any parametric form of the target kernel, nonparametric kernel learning offers a flexible scheme to utilize the information of the data, which may potentially characterize the data similarity better. The kernel methods using nonparametric kernels are referred to as nonparametric kernel methods. However, many nonparametric kernel methods are restricted to transductive learning, where the prediction function is defined only over the data points given beforehand. They have no straightforward extension for the out-of-sample data points, and thus cannot be applied to inductive learning. In this paper, we show how to make the nonparametric kernel methods applicable to inductive learning. The key problem of out-of-sample extension is how to extend the nonparametric kernel matrix to the corresponding kernel function. A regression approach in the hyper reproducing kernel Hilbert space is proposed to solve this problem. Empirical results indicate that the out-of-sample performance is comparable to the in-sample performance in most cases. Experiments on face recognition demonstrate the superiority of our nonparametric kernel method over the state-of-the-art parametric kernel methods.

Traditional nonnegative matrix factorization (NMF) is an unsupervised method for linear feature extraction. Recently, NMF with block strategy is shown to be able to extract more sparse and discriminative information of the images. To enhance the discriminative power of NMF, this paper proposes a block kernel nonnegative matrix factorization (BKNMF) based on the kernel theory and block technique. Kernel method is an effective way to model the nonlinear relations, which could help us to extract nonlinear features. Furthermore, we make use of the class label information to reduce the within-class distance for further improving the discriminative performance. We theoretically analyze the convergence of the proposed method. Three face databases, namely Yale, ORL and FERET databases, are chosen for evaluations. Compared with some state-of-the-art methods, experimental results show that our BKNMF approach achieves superior performance.

Automated guided vehicles (AGVs) have been largely used in manufacturing and supply chain management. With the development of Auto-ID technologies like radio frequency identification (RFID), AGVs' positioning could be enhanced. This paper demonstrates using magnetic field lines in the AGVs for precise coverage locating based on the errors' suppression positioning method. Dolph-Chebyshev antenna array is used to enable AGVs with more precise location implementation. It is observed that the far-field active RFID system positioning accuracy is higher, the movement is more stable, and the fluctuating rate is smaller.

The data from real world usually have nonlinear geometric structure, which are often assumed to lie on or close to a low-dimensional manifold in a high-dimensional space. How to detect this nonlinear geometric structure of the data is important for the learning algorithms. Recently, there has been a surge of interest in utilizing kernels to exploit the manifold structure of the data. Such kernels are called geometry-aware kernels and are widely used in the machine learning algorithms. The performance of these algorithms critically relies on the choice of the geometry-aware kernels. Intuitively, a good geometry-aware kernel should utilize additional information other than the geometric information. In many applications, it is required to compute the out-of-sample data directly. However, most of the geometry-aware kernel methods are restricted to the available data given beforehand, with no straightforward extension for out-of-sample data. In this paper, we propose a framework for more general geometry-aware kernel learning. The proposed framework integrates multiple sources of information and enables us to develop flexible and effective kernel matrices. Then, we theoretically show how the learned kernel matrices are extended to the corresponding kernel functions, in which the out-of-sample data can be computed directly. Under our framework, a novel family of geometry-aware kernels is developed. Especially, some existing geometry-aware kernels can be viewed as instances of our framework. The performance of the kernels is evaluated on dimensionality reduction, classification, and clustering tasks. The empirical results show that our kernels significantly improve the performance.

This paper investigates how to boost region-based image segmentation by inheriting the advantages of sparse representation and low-rank representation. A novel image segmentation model, called nonconvex regularization based sparse and low-rank coupling model, is presented for such a purpose. We aim at finding the optimal solution which is provided with sparse and low-rank simultaneously. This is achieved by relaxing sparse representation problem as L1/2 norm minimization other than the L1 norm minimization, while relaxing low-rank representation problem as the S1/2 norm minimization other than the nuclear norm minimization. This coupled model can be solved efficiently through the Augmented Lagrange Multiplier (ALM) method and half-threshold operator. Compared to the other state-of-the-art methods, the new method is better at capturing the global structure of the whole data, the robustness is better and the segmentation accuracy is also competitive. Experiments on two public image segmentation databases well validate the superiority of our method.

In this paper, visual attention spreading is formulated as a nonlocal diffusion equation. Different from other diffusion-based methods, a nonlocal diffusion tensor is introduced to consider both the diffusion strength and the diffusion direction. With the help of diffusion tensor, along with the principle direction, the diffusion has been suppressed to preserve the dissimilarity between the foreground and background, while in other directions, the diffusion has been boosted to combine the similar regions and highlight the salient object as a whole. Through a two-stages diffusion, the final saliency maps are obtained. Extensive quantitative or visual comparisons are performed on three widely used benchmark datasets, i.e. MSRA-ASD, MSRA-B and PASCAL-1500 datasets. Experimental results demonstrate the superior performance of our method.

In this paper, we propose a new fuzzy-based variational model that efficiently computes partitioning of speckled images, such as images obtained from Synthetic Aperture Radar (SAR). The model is derived by using the so-called maximizing a posteriori (MAP) estimation method. The novelties of the model are: (1) the Gamma distribution rather than the classical Gaussian distribution is used to model the gray intensities in each homogeneous region of the images (Gamma distribution function is better suited for speckled images); (2) an adaptive weighted regularization term with respect to a fuzzy membership function is designed to protect the segmentation results from degeneration (being over-smoothed). Compared with the classical total variation (TV) regularizer, the proposed regularization term has a sparser property. In addition, a new alternative direction iteration algorithm is proposed to solve the model. The algorithm is efficient since it integrates the split Bregman method and the Chambolle's projection method. Numerical examples are given to verify the efficiency of our model.

In this paper, the impact of the clustering exponent (α) on synchronizability, average shortest path length and navigability of small-world networks generated by one dimensional Kleinberg model is investigated. It could be seen from the analysis that the synchronizability becomes stronger as the clustering exponent decreases. And the navigability achieves peak at the neighborhood of α = 1, as well as the navigability becomes smaller as the clustering exponent increases. Moreover, the average path length of one dimensional Kleinberg small-world network decreases with respect to increasing clustering exponent. And this phenomenon is verified by numerical simulations on a network of Rossler oscillators. Then, it could be deduced from the phenomenon observed that compared with the low probabilities of longer distance of the edge-adding, the high probabilities of shorter distance of the edge-adding could achieve better synchronizability.