Transformer-based methods have shown impressive performance in low-level vision tasks, such as image super-resolution. However, we find that these networks can only utilize a limited spatial range of input information through attribution analysis. This implies the potential Transformer is still not fully exploited existing networks. In order to activate more pixels for better reconstruction, propose novel Hybrid Attention (HAT). It combines both channel attention and window-based...

Passive crossbar resistive random access memory (RRAM) arrays require select devices with nonlinear I-V characteristics to address the sneak-path problem. Here, we present a systematical analysis evaluate performance requirements of during read operation RRAM for proposed one-selector-one-resistor (1S1R) configuration serially connected selector/storage element. We found high selector current density is critical and nonlinearity (ON/OFF) requirement can be relaxed at present. Different...

Nanoscale resistive switching devices (memristive or memristors) have been studied for a number of applications ranging from non-volatile memory, logic to neuromorphic systems. However major challenge is address the potentially large variations in space and time these nanoscale devices. Here we show that metal-filament based memristive can be fully stochastic. While individual events are random, distribution probability well predicted controlled. Rather than trying force high probabilities...

The topics of visual and audio quality assessment (QA) have been widely researched for decades, yet nearly all this prior work has focused only on single-mode or signals. However, signals rarely are presented without accompanying audio, including heavy-bandwidth video streaming applications. Moreover, the distortions that may separately (or conjointly) afflict collectively shape user-perceived experience (QoE). This motivated us to conduct a subjective study (A/V) quality, which we then used...

A comprehensive analysis of write operations (SET and RESET) in a resistance-change memory (resistive random access memory) crossbar array is carried out. Three types resistive switching cells-nonlinear, rectifying-SET, rectifying-RESET-are compared with each other terms voltage delivery, current power consumption. Two different schemes, V/2 V/3, were considered, the scheme preferred due to much lower simple numerical method was developed that simulates entire flows node voltages within...

To enhance the visibility and usability of images captured in hazy conditions, many image dehazing algorithms (DHAs) have been proposed. With so DHAs, there is a need to evaluate compare these DHAs. Due lack reference haze-free images, DHAs are generally evaluated qualitatively using real images. But it possible perform quantitative evaluation synthetic since available full-reference (FR) quality assessment (IQA) measures can be utilized. In this paper, we follow strategy study DHA...

This paper proposes a novel reversible image data hiding scheme over encrypted domain. Data embedding is achieved through public key modulation mechanism, in which access to the secret encryption not needed. At decoder side, powerful two-class SVM classifier designed distinguish and nonencrypted patches, allowing us jointly decode embedded message original signal. Compared with state-of-the-art methods, proposed approach provides higher capacity able perfectly reconstruct as well message....

Audio information has been bypassed by most of current visual attention prediction studies. However, sound could have influence on and such widely investigated proofed many psychological In this paper, we propose a novel multi-modal saliency (MMS) model for videos containing scenes with high audio-visual correspondence. scenes, humans tend to be attracted the sources it is also possible localize via cross-modal analysis. Specifically, first detect spatial temporal maps from modality using...

In many practical scenarios, image encryption has to be conducted prior compression. This led the problem of how design a pair and compression algorithms such that compressing encrypted images can still efficiently performed. this paper, we highly efficient encryption-then-compression (ETC) system, where both lossless lossy are considered. The proposed scheme operated in prediction error domain is shown able provide reasonably high level security. We also demonstrate an arithmetic...

Copy-move forgery is one of the most commonly used manipulations for tampering digital images. Keypoint-based detection methods have been reported to be very effective in revealing copy-move evidence due their robustness against various attacks, such as large-scale geometric transformations. However, these fail handle cases when forgeries only involve small or smooth regions, where number keypoints limited. To tackle this challenge, we propose a fast and algorithm through hierarchical...

The increasing abuse of image editing software causes the authenticity digital images questionable. Meanwhile, widespread availability online social networks (OSNs) makes them dominant channels for transmitting forged to report fake news, propagate rumors, etc. Unfortunately, various lossy operations, e.g., compression and resizing, adopted by OSNs impose great challenges implementing robust forgery detection. To fight against OSN-shared forgeries, in this work, a novel training scheme is...

Owning to the recorded light ray distributions, field contains much richer information and provides possibilities of some enlightening applications, it has becoming more popular. To facilitate relevant many processing techniques have been proposed recently. These operations also bring loss visual quality, thus there is need a quality metric quantify loss. reduce complexity resource consumption, fields are generally sparsely sampled, compressed, finally reconstructed displayed users. We...

The state-of-the-art graph wavelet decomposition was constructed by maximum spanning tree (MST)-based downsampling and two-channel filter banks. In this work, we first show that: 1) the existing MST-based could become unbalanced, i.e., sampling rate is far from 1/2, which eventually leads to low representation efficiency of decomposition; 2) not only low-pass components, but also some high-pass ones can be decomposed potentially achieve better performance. Based on these observations,...

To solve the "big data" problems that are hindered by Von Neumann bottleneck and semiconductor device scaling limitation, a new efficient in-memory computing architecture based on crossbar array is developed. The corresponding basic operation principles design rules proposed verified using emerging nonvolatile devices such as very low-power resistive random access memory (RRAM). prove architecture, we demonstrate parallel 1-bit full adder (FA) both experiment simulation. A 4-bit multiplier...

Through exploiting the image nonlocal self-similarity (NSS) prior by clustering similar patches to construct patch groups, recent studies have revealed that structural sparse representation (SSR) models can achieve promising performance in various restoration tasks. However, most existing SSR methods only exploit NSS from input degraded (internal) image, and few utilize external clean corpus; how jointly priors of internal corpus is still an open problem. In this article, we propose a novel...

To resolve the sneak leakage problem and reduce power consumption in crossbar RRAM arrays, a Cu/Al <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> O xmlns:xlink="http://www.w3.org/1999/xlink">3</sub> /aSi/Ta cell with self-rectifying characteristics is developed. The exhibits low operating current (~nA), high ON/OFF ratios (>100×), pronounced nonlinearity. use of low-programming-current elements avoids current-driving capability...

Sparse coding has achieved a great success in various image processing tasks. However, benchmark to measure the sparsity of patch/group is missing since sparse essentially an NP-hard problem. This work attempts fill gap from perspective rank minimization. We firstly design adaptive dictionary bridge between group-based (GSC) and Then, we show that under designed dictionary, GSC minimization problems are equivalent, therefore coefficients each patch group can be measured by estimating...

In this paper, we propose a novel approach to the rank minimization problem, termed residual constraint (RRC) model. Different from existing low-rank based approaches, such as well-known nuclear norm (NNM) and weighted (WNNM), which estimate underlying matrix directly corrupted observations, progressively approximate via minimizing residual. Through integrating image nonlocal self-similarity (NSS) prior with proposed RRC model, apply it restoration tasks, including denoising compression...

Group sparse representation (GSR) has made great strides in image restoration producing superior performance, realized through employing a powerful mechanism to integrate the local sparsity and nonlocal self-similarity of images. However, due some form degradation (e.g., noise, down-sampling or pixels missing), traditional GSR models may fail faithfully estimate each group an image, thus resulting distorted reconstruction original image. This motivates us design simple yet effective model...

Sparse representation has achieved great success in various image processing and computer vision tasks. For processing, typical patch-based sparse (PSR) models usually tend to generate undesirable visual artifacts, while group-based (GSR) lean produce over-smooth effects. In this paper, we propose a new model, termed joint patch-group based (JPG-SR). Compared with existing models, the proposed JPG-SR provides an effective mechanism integrate local sparsity nonlocal self-similarity of images....

Deep learning (DL) has demonstrated its powerful capabilities in the field of image inpainting, which could produce visually plausible results. Meanwhile, malicious use advanced inpainting tools (e.g. removing key objects to report fake news, erasing visible copyright watermarks, etc.) led increasing threats reliability data. To fight against forgeries (not only DL-based but also traditional ones), this work, we propose a novel end-to-end Image Inpainting Detection Network (IID-Net), detect...

By recording the whole scene around capturer, virtual reality (VR) techniques can provide viewers sense of presence. To a satisfactory quality experience, there should be at least 60 pixels per degree, so resolution panoramas reach 21600 × 10800. The huge amount data will put great demands on processing and transmission. However, when exploring in environment, only perceive content current field view (FOV). Therefore if we predict head eye movements which are important behaviors viewer, more...