Segmenting unknown domains using a model trained in the source domain still faces challenges. Although some approaches tried to resolve the problem through various data generation and network architecture designs, they cannot achieve satisfactory segmentation results compared with single domain segmentation of consistent data distribution. Therefore, we propose a data augmentation method based on amplitude perturbation to expand the distribution of data types, thereby covering target data. A feature suppression strategy is proposed to reduce the network's over-reliance on important features of the source domain data to improve generalization performance. In addition, we design a luminance contrast consistency (LCC) learning module to harmonize the data styles between different domains and a multiscale convolutional attention (MSCA) module to enhance the network's perception of small target objects and improve the segmentation performance of the model, which further improves segmentation performance. Our method achieves the state-of-the-art (SOTA) results on two public datasets of ATLAS2.0 and Prostate. The code is available at https://github.com/butterflyGN/DGSFTAFS.

Image hiding is often referred to as steganography, which aims to hide a secret image in a cover image of the same resolution. Many steganography models are based on generative adversarial networks (GANs) and variational autoencoders (VAEs). However, most existing models suffer from mode collapse. Mode collapse will lead to an information imbalance between the cover and secret images in the stego image and further affect the subsequent extraction. To address these challenges, this paper proposes StegOT, an autoencoder-based steganography model incorporating optimal transport theory. We designed the multiple channel optimal transport (MCOT) module to transform the feature distribution, which exhibits multiple peaks, into a single peak to achieve the trade-off of information. Experiments demonstrate that we not only achieve a trade-off between the cover and secret images but also enhance the quality of both the stego and recovery images. The source code will be released on https://github.com/Rss1124/StegOT.

Detecting surface defects in industrial products is crucial for ensuring quality control in manufacturing. Traditional methods face challenges due to the diversity of defect types, the small and ambiguous nature of defects, and the high cost of labeled data. Unsupervised and semi-supervised methods can reduce labeling costs but often fail to meet industrial accuracy requirements. In this paper, we propose an Enhanced Hybrid Supervised Framework (EHSF) designed to improve defect detection accuracy in complex industrial scenarios with fewer labeled samples. The framework incorporates an Adaptive Cross-Scale Feature Enhancement Module (ACFEM) based on Selective Feature Aggregation (SFA), which addresses the limitations of single-level feature representations and significantly enhances the detection of multiscale defects. Additionally, we introduce a novel Dynamic Feature Calibration Network (DFCNet) that synergistically combines global contextual information and local details through dynamic feature calibration and adaptive global semantic enhancement mechanisms. The proposed approach is validated on the DAGM benchmark and three real-world industrial datasets (Kolektor Surface Defect Dataset, Kolektor Surface Defect Dataset2, and Severstal Steel). Experimental results demonstrate that our method outperforms existing techniques by reducing reliance on weakly supervised fine annotations while achieving superior detection accuracy, particularly under weak supervision.

Abstract An intelligent retrievable object‐tracking system assists users in quickly and accurately locating lost objects. However, challenges such as real‐time processing on edge devices, low image resolution, and small‐object detection significantly impact the accuracy and efficiency of video‐stream‐based systems, especially in indoor home environments. To overcome these limitations, a novel real‐time intelligent retrievable object‐tracking system is designed. The system incorporates a retrievable object‐tracking algorithm that combines DeepSORT and sliding window techniques to enhance tracking capabilities. Additionally, the YOLOv7‐small‐scale model is proposed for small‐object detection, integrating a specialized detection layer and the convolutional batch normalization LeakyReLU spatial‐depth convolution module to enhance feature capture for small objects. TensorRT and INT8 quantization are used for inference acceleration on edge devices, doubling the frames per second. Experiments on a Jetson Nano (4 GB) using YOLOv7‐small‐scale show an 8.9% improvement in recognition accuracy over YOLOv7‐tiny in video stream processing. This advancement significantly boosts the system's performance in efficiently and accurately locating lost objects in indoor home settings.

Semi-supervised learning for medical image segmentation presents a unique challenge of efficiently using limited labeled data while leveraging abundant unlabeled data. Despite advancements, existing methods often do not fully exploit the potential of the unlabeled data for enhancing model robustness and accuracy. In this paper, we introduce CrossMatch, a novel framework that integrates knowledge distillation with dual perturbation strategies, image-level and feature-level, to improve the model's learning from both labeled and unlabeled data. CrossMatch employs multiple encoders and decoders to generate diverse data streams, which undergo self-knowledge distillation to enhance the consistency and reliability of predictions across varied perturbations. Our method significantly surpasses other state-of-the-art techniques in standard benchmarks by effectively minimizing the gap between training on labeled and unlabeled data and improving edge accuracy and generalization in medical image segmentation. The efficacy of CrossMatch is demonstrated through extensive experimental validations, showing remarkable performance improvements without increasing computational costs.

Medical image processing plays an important role in the interaction of real world and metaverse for healthcare. Self-supervised denoising based on sparse coding methods, without any prerequisite on large-scale training samples, has been attracting extensive attention for medical image processing. Whereas, existing self-supervised methods suffer from poor performance and low efficiency. In this paper, to achieve state-of-the-art denoising performance on the one hand, we present a self-supervised sparse coding method, named the weighted iterative shrinkage thresholding algorithm (WISTA). It does not rely on noisy-clean ground-truth image pairs to learn from only a single noisy image. On the other hand, to further improve denoising efficiency, we unfold the WISTA to construct a deep neural network (DNN) structured WISTA, named WISTA-Net. Specifically, in WISTA, motivated by the merit of the$l_{p}$-norm, WISTA-Net has better denoising performance than the classical orthogonal matching pursuit (OMP) algorithm and the ISTA. Moreover, leveraging the high-efficiency of DNN structure in parameter updating, WISTA-Net outperforms the compared methods in denoising efficiency. In detail, for a 256 by 256 noisy image, the running time of WISTA-Net is 4.72 s on the CPU, which is much faster than WISTA, OMP, and ISTA by 32.88 s, 13.06 s, and 6.17 s, respectively.

Abstract Video summarization offers a summary version that conveys the primary information of a longer video. The main challenges of video summarization are related to keyframe extraction and saliency mapping. Thus, this work proposes a sparse coding model for keyframe extraction and saliency mapping applications. Specifically, the minimax concave penalty (MCP) is utilized as a sparse regularization scheme and the regularized non‐convex MCP problem is solved by decomposing MCP into two convex functions and the convex function's algorithm difference is relied on to solve the resulting sub‐problems. The experimental results demonstrate higher compressed keyframes and saliency maps than current state‐of‐the‐art algorithms. In particular, the model attains a lower summary length of 34% and 19% compared to sparse modeling representation selection (SMRS) and sparse modeling using the determinant sparsity measure (SC‐det), respectively. In addition, the developed scheme has a shorter computation time, requiring 82% and 33% less time than the ITTI and the dense and sparse reconstruction (DSR) methods.

Abstract Acute ischemic stroke is a common threat to human health and may obtain timely treatment by fast localizing and quantitatively evaluating the lesions. Most CNN‐based methods try to segment and measure the lesions, however, they require a training on a large number of labeled subjects that are labor‐intensive and time‐consuming to obtain. In this paper, a method is proposed that can combine limited labeled subjects with abundant unlabeled subjects to alleviate the problem. The proposed method consists of two stages: stepwise learning process and segmentation process. Stepwise learning is used to obtain the pretrained encoder. The pretrained encoder and the proposed decoder are connected into a new end‐to‐end segmentation network, which is retrained on the labeled subjects in the segmentation process. By using 5 labeled subjects and 79 unlabeled subjects, the proposed method achieves a mean dice coefficient of 0.6630.205, a mean average symmetric surface distance (ASSD) of 2.17 mm and a mean 95 percentile Hausdorff distance (HD) of 18.38 mm on a clinical MR dataset with 179 subjects. More importantly, it achieves lesion‐wise F 1 score of 0.857 and a subject‐wise detection rate of 0.966.

Device-free localization (DFL) locates targets without equipping with wireless devices or tag under the Internet-of-Things (IoT) architectures. As an emerging technology, DFL has spawned extensive applications in the IoT environment, such as intrusion detection, mobile robot localization, and location-based services. Current DFL-related machine learning (ML) algorithms still suffer from low localization accuracy and weak dependability/robustness because the group structure has not been considered in their location estimation, which leads to an undependable process. To overcome these challenges, we propose in this work a dependable block-sparse scheme by particularly considering the group structure of signals. An accurate and robust ML algorithm named block-sparse coding with the proximal operator (BSCPO) is proposed for DFL. In addition, a severe Gaussian noise is added in the original sensing signals for preserving network-related privacy as well as improving the dependability of the model. The real-world data-driven experimental results show that the proposed BSCPO achieves robust localization and signal-recovery performance even under severely noisy conditions and outperforms state-of-the-art DFL methods. For single-target localization, BSCPO retains high accuracy when the signal-to-noise ratio exceeds -10 dB. BSCPO is also able to localize accurately under most multitarget localization test cases.

Gaze prediction is a significant approach for processing a large amount of incoming visual information of videos. Recent gaze prediction algorithms often employ sparse models with the assumption that every superpixel in the video frames can be represented as linear combinations of a few salient superpixels. However, they are not actuated enough because of the insufficient knowledge that video signals contain a non-negative request. Hence, we develop a novel gaze prediction based on an inverse sparse coding framework with a determinant sparse measure. By introducing this sparse measure, the solutions are non-negative and sparser than conventional sparse constraints. However, the proposed optimization problem becomes nonconvex, which is difficult to solve. To efficiently address the corresponding nonconvex optimization problem, we propose a novel algorithm based on the difference in convex function programming, which can yield the global solutions. Experimental results indicate the improved accuracy of the proposed approach compared with state-of-the-art algorithms.

The user-centric cloud radio access network (C-RAN) is promising for significantly reducing the channel training overhead because only the intra-cluster channel state information (CSI) is required. However, the inter-cluster interference may degrade the network performance. To address this problem, we present a novel framework for the uplink of user-centric C-RANs where the interference cancellation is posed as a system of bi-unknown vectors linear equations. To solve this unusual system of equations, we propose a quasi-least squares (QLS) algorithm and analyze its robustness by exploiting the random matrix theory. We reveal the fact that QLS is very sensitive to the channel estimation error due to involving the inverse of an ill-conditioned matrix. It is well known that truncated singular value decomposition (TSVD) is an effective regularization scheme that can mitigate this ill-conditioning effect. Accordingly, we employ TSVD to further improve the robustness of QLS against the imperfect channel estimation. In addition, since the performance of the TSVD based algorithm strongly depends on the truncation parameter, a parameter-choice method using the constant modulus (CM) feature is also provided. Finally, simulation results are presented to examine the effectiveness and robustness of the proposed method.

Fault detection is a fundamental requirement for Industrial Internet of Things (IIoT), such as the process industry. This article first reviews the recent studies focusing on applying the fault detection techniques to the IIoT networks. However, we find that numerous studies focus on the resource utilization and workload allocation. The fault detection toward IIoT facilities is still in its immature stage because the existing approaches are not accurate enough for the stringent fault detection in IIoT networks. To this end, we present a novel algorithm, named Gaussian Bernoulli restricted Boltzmann machines (GBRBMs)-based deep neural network (DNN), to transform the fault detection into a classification problem. The real trace-driven experiments show that the proposed scheme outperforms other baseline machine learning methods. We anticipate that this article can inspire blooming studies on the related topics of smart IIoT networks.

Device-free localization (DFL), as an emerging technology that locates targets without any attached devices via wireless sensor networks, has spawned extensive applications in the Internet of Things (IoT) field. For DFL, a key problem is how to extract significant features to characterize raw signals with different patterns associated with different locations. To address this problem, in this paper, the DFL problem is formulated as an image classification problem. Moreover, we design a three-layer convolutional autoencoder (CAE) neural network to perform unsupervised feature extraction from raw signals followed by supervised fine-tuning for classification. The CAE combines the advantages of a convolutional neural network (CNN) and a deep autoencoder (AE) in the feature learning and signals reconstruction, which is expected to achieve good performance for DFL. The experimental results show that the proposed approach can achieve a high localization accuracy rate of 100% for a reasonable grid size on the raw real-world data, i.e., the collected raw data without added Gaussian noise, and is robust to noisy data with a signal-to-noise ratio greater than -5 dB. Additionally, its time cost for the classification of a single activity is 4 ms, which is fast enough for the IoT applications. The proposed approach outperforms the deep CNN and AE in terms of localization accuracy and robust ability against noise.

This paper is for addressing the nonnegative sparse representation problem, i.e. to represent a nonnegative matrix as an over complete nonnegative dictionary times a nonnegative coefficient matrix. Many data in the real world can be represented sparsely by combinations of typical features. Moreover, large family of nonnegative data, such as image pixels, word frequency, power spectrum etc., are in great demand for engineering problems. Nonnegative Sparse Representation (NSR) is attractive to nonnegative data analysis. In this study, Overcomplete dictionary learning for NSR with an ℓp-norm (0pp-norm is expected for leading a sparser solution than ℓ1-norm. An ℓp-norm is non-convex, then ℓp-norm was approximated by weighted ℓ1-norm for convex optimization. We performed experiments about recovering accuracies of dictionary and coefficients. The recovering ratio is evaluated for various sparse levels of coefficients. The average of recovery ratios by proposed method for each sparse level were higher compared with an unweighted ℓ1-norm. It improved +15.37% at most. We have shown that the proposed method also has advantages in recovering sparseness, ℓ2relative error and support distance of coefficients etc.

As an emerging technology, device-free localization (DFL), using radio frequency (RF) sensor networks to detect targets who do not carry any attached devices, has spawned extensive applications. Many existing works formulate DFL as a classification problem, and a key problem is how to extract discriminative features to characterize the raw wireless signal. In this paper, we present an autoencoder-based deep neural network for feature extraction, moreover, multiple Gaussian Bernoulli restricted Boltzmann machines (GBRBMs) are utilized for pre-training and dimension reduction. Experiment results show that this method of GBRBM-based autoencoder (GBRBM-AE) can achieve a high accuracy and efficient performance, which outperforms the conventional autoencoder. When the dimensions of input data are reduced from 784 to 20 dims, our algorithm can maintain a high accuracy of 97.1% and is robust to noise with SNR = 5dB.

In analysis dictionary learning, the learned dictionary may contain similar atoms, leading to a degenerate dictionary. To address this problem, we propose a novel incoherent analysis dictionary learning algorithm with the ℓ1-norm for sparsity and simultaneously with the coherence penalty. The whole problem is convex but nonsmooth due to the sparsity regularizer and the coherence penalty. Hence, the proximal operator is introduced to conquer the nonsmoothness in the sparsity regularizer and in the coherence penalty. The alternating minimization is sequentially solved for each row of the analysis dictionary and for each row of the analysis matrix in the same manner. According to our analysis and simulation study, the main advantages of the proposed algorithm are its greater efficiency in learning and its higher convergence rate than state-of-the-art algorithms.

We present a fast, efficient algorithm for learning an overcomplete dictionary for sparse representation of signals. The whole problem is considered as a minimization of the approximation error function with a coherence penalty for the dictionary atoms and with the sparsity regularization of the coefficient matrix. Because the problem is nonconvex and nonsmooth, this minimization problem cannot be solved efficiently by an ordinary optimization method. We propose a decomposition scheme and an alternating optimization that can turn the problem into a set of minimizations of piecewise quadratic and univariate subproblems, each of which is a single variable vector problem, of either one dictionary atom or one coefficient vector. Although the subproblems are still nonsmooth, remarkably they become much simpler so that we can find a closed-form solution by introducing a proximal operator. This leads to an efficient algorithm for sparse representation. To our knowledge, applying the proximal operator to the problem with an incoherence term and obtaining the optimal dictionary atoms in closed form with a proximal operator technique have not previously been studied. The main advantages of the proposed algorithm are that, as suggested by our analysis and simulation study, it has lower computational complexity and a higher convergence rate than state-of-the-art algorithms. In addition, for real applications, it shows good performance and significant reductions in computational time.

The problem of nonnegative blind source separation (NBSS) is addressed in this paper, where both the sources and the mixing matrix are nonnegative. Because many real-world signals are sparse, we deal with NBSS by sparse component analysis. First, a determinant-based sparseness measure, named D-measure, is introduced to gauge the temporal and spatial sparseness of signals. Based on this measure, a new NBSS model is derived, and an iterative sparseness maximization (ISM) approach is proposed to solve this model. In the ISM approach, the NBSS problem can be cast into row-to-row optimizations with respect to the unmixing matrix, and then the quadratic programming (QP) technique is used to optimize each row. Furthermore, we analyze the source identifiability and the computational complexity of the proposed ISM-QP method. The new method requires relatively weak conditions on the sources and the mixing matrix, has high computational efficiency, and is easy to implement. Simulation results demonstrate the effectiveness of our method.

Nonnegative matrix factorization (NMF) is a widely used method for blind spectral unmixing (SU), which aims at obtaining the endmembers and corresponding fractional abundances, knowing only the collected mixing spectral data. It is noted that the abundance may be sparse (i.e., the endmembers may be with sparse distributions) and sparse NMF tends to lead to a unique result, so it is intuitive and meaningful to constrain NMF with sparseness for solving SU. However, due to the abundance sum-to-one constraint in SU, the traditional sparseness measured by L0/L1-norm is not an effective constraint any more. A novel measure (termed as S-measure) of sparseness using higher order norms of the signal vector is proposed in this paper. It features the physical significance. By using the S-measure constraint (SMC), a gradient-based sparse NMF algorithm (termed as NMF-SMC) is proposed for solving the SU problem, where the learning rate is adaptively selected, and the endmembers and abundances are simultaneously estimated. In the proposed NMF-SMC, there is no pure index assumption and no need to know the exact sparseness degree of the abundance in prior. Yet, it does not require the preprocessing of dimension reduction in which some useful information may be lost. Experiments based on synthetic mixtures and real-world images collected by AVIRIS and HYDICE sensors are performed to evaluate the validity of the proposed method.

Blind source separation (BSS) has been widely discussed in many real applications. Recently, under the assumption that both of the sources and the mixing matrix are nonnegative, Wang develop an amazing BSS method by using volume maximization. However, the algorithm that they have proposed can guarantee the nonnegativities of the sources only, but cannot obtain a nonnegative mixing matrix necessarily. In this letter, by introducing additional constraints, a method for fully nonnegative constrained iterative volume maximization (FNCIVM) is proposed. The result is with more interpretation, while the algorithm is based on solving a single linear programming problem. Numerical experiments with synthetic signals and real-world images are performed, which show the effectiveness of the proposed method.

Convolutive blind source separation (CBSS) that exploits the sparsity of source signals in the frequency domain is addressed in this paper. We assume the sources follow complex Laplacian-like distribution for complex random variable, in which the real part and imaginary part of complex-valued source signals are not necessarily independent. Based on the maximum a posteriori (MAP) criterion, we propose a novel natural gradient method for complex sparse representation. Moreover, a new CBSS method is further developed based on complex sparse representation. The developed CBSS algorithm works in the frequency domain. Here, we assume that the source signals are sufficiently sparse in the frequency domain. If the sources are sufficiently sparse in the frequency domain and the filter length of mixing channels is relatively small and can be estimated, we can even achieve underdetermined CBSS. We illustrate the validity and performance of the proposed learning algorithm by several simulation examples.

We propose an approach for real-time blind source separation (BSS), in which the observations are linear convolutive mixtures of statistically independent acoustic sources. A recursive least square (RLS)-like strategy is devised for real-time BSS processing. A normal equation is further introduced as an expression between the separation matrix and the correlation matrix of observations. We recursively estimate the correlation matrix and explicitly, rather than stochastically, solve the normal equation to obtain the separation matrix. As an example of application, the approach has been applied to a BSS problem where the separation criterion is based on the second-order statistics and the non-stationarity of signals in the frequency domain. In this way, we realise a novel BSS algorithm, called exponentially weighted recursive BSS algorithm. The simulation and experimental results showed an improved separation and a superior convergence rate of the proposed algorithm over that of the gradient algorithm. Moreover, this algorithm can converge to a much lower cost value than that of the gradient algorithm.