Sparse representation models code an image patch as a linear combination of few atoms chosen out from over-complete dictionary, and they have shown promising results in various restoration applications. However, due to the degradation observed (e.g., noisy, blurred, and/or down-sampled), sparse representations by conventional may not be accurate enough for faithful reconstruction original image. To improve performance representation-based restoration, this paper concept coding noise is...

As a powerful statistical image modeling technique, sparse representation has been successfully used in various restoration applications. The success of owes to the development <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">l</i> <sub xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -norm optimization techniques and fact that natural images are intrinsically some domains. quality largely depends on whether employed domain can represent well...

Simultaneous sparse coding (SSC) or nonlocal image representation has shown great potential in various low-level vision tasks, leading to several state-of-the-art restoration techniques, including BM3D and LSSC. However, it still lacks a physically plausible explanation about why SSC is better model than conventional for the class of natural images. Meanwhile, problem sparsity optimization, especially when tangled with dictionary learning, computationally difficult solve. In this paper, we...

Sparsity has been widely exploited for exact reconstruction of a signal from small number random measurements. Recent advances have suggested that structured or group sparsity often leads to more powerful techniques in various compressed sensing (CS) studies. In this paper, we propose nonlocal low-rank regularization (NLR) approach toward exploiting and explore its application into CS both photographic MRI images. We also the use nonconvex log det ( X) as smooth surrogate function rank...

Hyperspectral imaging has many applications from agriculture and astronomy to surveillance mineralogy. However, it is often challenging obtain high-resolution (HR) hyperspectral images using existing techniques due various hardware limitations. In this paper, we propose a new image super-resolution method low-resolution (LR) HR reference of the same scene. The estimation formulated as joint dictionary sparse codes based on prior knowledge spatial-spectral sparsity image. representing...

Where does the sparsity in image signals come from? Local and nonlocal models have supplied complementary views toward regularity natural images - former attempts to construct or learn a dictionary of basis functions that promotes sparsity; while latter connects with self-similarity source by clustering. In this paper, we present variational framework for unifying above two propose new denoising algorithm built upon clustering-based sparse representation (CSR). Inspired success l <sub...

Deep neural networks (DNNs) have shown very promising results for various image restoration (IR) tasks. However, the design of network architectures remains a major challenging achieving further improvements. While most existing DNN-based methods solve IR problems by directly mapping low quality images to desirable high-quality images, observation models characterizing degradation processes been largely ignored. In this paper, we first propose denoising-based algorithm, whose iterative steps...

Sparse representation is proven to be a promising approach image super-resolution, where the low-resolution (LR) usually modeled as down-sampled version of its high-resolution (HR) counterpart after blurring. When blurring kernel Dirac delta function, i.e., LR directly from HR without blurring, super-resolution problem becomes an interpolation problem. In such cases, however, conventional sparse models (SRM) become less effective, because data fidelity term fails constrain local structures....

In this paper, a reflective metasurface is designed, fabricated, and experimentally demonstrated to generate an orbital angular momentum (OAM) vortex wave in radio frequency domain. Theoretical formula of phase-shift distribution deduced used design the producing waves. The prototype practical configuration measured validate theoretical analysis at 5.8 GHz. simulated experimental results verify that waves with different OAM mode numbers can be flexibly generated by using sub-wavelength...

Recently, increasing interest has been drawn in exploiting deep convolutional neural networks (DCNNs) for no-reference image quality assessment (NR-IQA). Despite of the notable success achieved, there is a broad consensus that training DCNNs heavily relies on massive annotated data. Unfortunately, IQA typical small sample problem. Therefore, most existing DCNN-based metrics operate based pre-trained networks. However, these are not designed task, leading to generalization problem when...

In this paper, an electromagnetic metasurface is designed, fabricated, and experimentally demonstrated to generate multiple orbital angular momentum (OAM) vortex beams in radio frequency domain. Theoretical formula of compensated phase-shift distribution deduced used design the produce waves different directions with OAM modes. The prototype a practical configuration square-patch measured validate theoretical analysis at 5.8 GHz. simulated experimental results verify that can be...

This paper proposes a novel sparse representation model called centralized (CSR) for image restoration tasks. In order faithful reconstruction, it is expected that the coding coefficients of degraded should be as close possible to those unknown original with given dictionary. However, since available data are (noisy, blurred and/or down-sampled) versions image, often not accurate enough if only local sparsity considered, in many existing models. To make more accurate, constraint introduced...

Existing communication systems are mainly built based on Shannon's information theory, which deliberately ignores the semantic aspects of communication. The recent iteration wireless technology, 5G and beyond, promises to support a plethora services enabled by carefully tailored network capabilities contents, requirements, as well semantics. This has sparked significant interest in communication, novel paradigm that involves meaning messages In this article, we first review classic...

Small objects detection is a challenging task in computer vision due to its limited resolution and information. In order solve this problem, the majority of existing methods sacrifice speed for improvement accuracy. paper, we aim detect small at fast speed, using best object detector Single Shot Multibox Detector (SSD) with respect accuracy-vs-speed trade-off as base architecture. We propose multi-level feature fusion method introducing contextual information SSD, improve accuracy objects....

The deep convolutional neural network (CNN) has achieved great success in image recognition. Many quality assessment (IQA) methods directly use recognition-oriented CNN for prediction. However, the properties of IQA task is different from recognition task. Image should be sensitive to visual content and robust distortion, while both distortion content. In this paper, an IQA-oriented method developed blind (BIQA), which can efficiently represent degradation. large-data driven, sizes existing...

The trade-off between spatial and spectral resolution is one of the fundamental issues in hyperspectral images (HSI). Given challenges directly acquiring high-resolution (HR-HSI), a compromised solution to fuse pair images: has (HR) domain but low-resolution (LR) spectral-domain other vice versa. Model-based image fusion methods including pan-sharpening aim at reconstructing HR-HSI by solving manually designed objective functions. However, such hand-crafted prior often leads inevitable...

In coded aperture snapshot spectral imaging (CASSI) system, the real-world hyperspectral image (HSI) can be reconstructed from captured compressive in a snapshot. Model-based HSI reconstruction methods employed hand-crafted priors to solve problem, but most of which achieved limited success due poor representation capability these priors. Deep learning based mappings between images and HSIs directly much better results. Yet, it is nontrivial design powerful deep network heuristically for...

Previous electroencephalogram (EEG) emotion recognition relies on single-task learning, which may lead to overfitting and learned features lacking generalization. In this paper, a graph-based multi-task self-supervised learning model (GMSS) for EEG is proposed. GMSS has the ability learn more general representations by integrating multiple tasks, including spatial frequency jigsaw puzzle contrastive tasks. By from tasks simultaneously, can find representation that captures all of thereby...

Convolutional neural networks (CNNs) have been successfully applied to the single target tracking task in recent years. Generally, training a deep CNN model requires numerous labeled samples, and number quality of these samples directly affect representational capability trained model. However, this approach is restrictive practice, because manually labeling such large time-consuming prohibitively expensive. In article, we propose an active learning method for visual tracking, which selects...

Coded aperture snapshot spectral imaging (CASSI) provides an efficient mechanism for recovering 3D data from a single 2D measurement. However, since the reconstruction problem is severely underdetermined, quality of recovered usually limited. In this paper we propose novel dual-camera design to improve performance CASSI while maintaining its advantage. Specifically, beam splitter placed in front objective lens CASSI, which allows same scene be simultaneously captured by grayscale camera....

Leveraging the compressive sensing (CS) theory, coded aperture snapshot spectral imaging (CASSI) provides an efficient solution to recover 3D hyperspectral data from a 2D measurement. The dual-camera design of CASSI, by adding uncoded panchromatic measurement, enhances reconstruction fidelity while maintaining advantage. In this paper, we propose adaptive nonlocal sparse representation (ANSR) model boost performance (DCCHI). Specifically, CS problem is formulated as cube based make full use...

Reduced-reference (RR) image quality assessment (IQA) aims to use less data about the reference and achieve higher evaluation accuracy. Recent research on brain theory suggests that human visual system (HVS) actively predicts primary information tries avoid residual uncertainty for perception understanding. Therefore, perceptual relies fidelities of uncertainty. In this paper, we propose a novel RR IQA index based fidelity. We advocate distortions mainly disturb understanding, change comfort...