In order to resist the adverse effect of viewpoint variations, we design quadruple directional deep learning networks extract features (QD-DLF) vehicle images for improving re-identification performance. The are similar overall architecture, including same basic architecture but different feature pooling layers. Specifically, that is a shortly and densely connected convolutional neural network utilized maps an input square image in first stage. Then, utilize layers, i.e., horizontal average...

Transformers are more and popular in computer vision, which treat an image as a sequence of patches learn robust global features from the sequence. However, pure transformers not entirely suitable for vehicle re-identification because requires both discriminative local features. For that, graph interactive transformer (GiT) is proposed this paper. In macro view, list GiT blocks stacked to build model, where graphs extract within among patches. micro status, bringing effective cooperation...

Strip steel surface defect detection is a critical step in the production field of industry and vital guarantee to improve quality strip production. However, due poor contrast images, diversity types, scales, texture structures, irregular distribution defects, it difficult achieve rapid accurate defects with existing methods. In this article, network for based on deformable convolution attention mechanism (DCAM-Net) proposed. First, we introduce limited adaptive histogram equalization...

In this paper, an accurate full-reference image quality assessment (IQA) model developed for assessing screen content images (SCIs), called the edge similarity (ESIM), is proposed. It inspired by fact that human visual system (HVS) highly sensitive to edges are often encountered in SCIs; therefore, essential features extracted and exploited conducting IQA SCIs. The key novelty of proposed ESIM lies extraction use three salient features-i.e., contrast, width, direction. first two attributes...

The intra-mode and inter-mode predictions have been made available in H.264/AVC for effectively improving coding efficiency. However, exhaustively checking all the prediction modes identifying best one (commonly referred to as <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">exhaustive</i> xmlns:xlink="http://www.w3.org/1999/xlink">mode</i> xmlns:xlink="http://www.w3.org/1999/xlink">decision</i> ) greatly increases computational complexity. In this...

In this paper, an accurate and efficient full-reference image quality assessment (IQA) model using the extracted Gabor features, called feature-based (GFM), is proposed for conducting objective evaluation of screen content images (SCIs). It well-known that filters are highly consistent with response human visual system (HVS), HVS sensitive to edge information. Based on these facts, imaginary part filter has odd symmetry yields detection exploited luminance reference distorted SCI extracting...

The emergence of 3D point clouds (3DPCs) is promoting the rapid development immersive communication, autonomous driving, and so on. Due to huge data volume, compression 3DPCs becoming more attractive. We propose a novel efficient color attribute method for static 3DPCs. First, 3DPC partitioned into several sub-point by distribution analysis. Each cloud then decomposed lot blocks an improved k-d tree-based decomposition algorithm. Afterwards, virtual adaptive sampling-based sparse...

This paper explores the problem of hyperspectral image (HSI) super-resolution that merges a low resolution HSI (LR-HSI) and high multispectral (HR-MSI). The cross-modality distribution spatial spectral information makes challenging. Inspired by classic wavelet decomposition-based fusion, we propose novel \textit{lightweight} deep neural network-based framework, namely progressive zero-centric residual network (PZRes-Net), to address this efficiently effectively. Specifically, PZRes-Net...

With the vigorous development of Industry 4.0, industrial Big Data has turned into core element Industrial Internet Things. As one most fundamental and indispensable components in cyber-physical systems (CPS), intelligent anomaly detection is still an essential challenging issue. However, with network, there may exist unknown types attacks, which are difficult to collect. Facing one-class intrusion scenario that collected training data only includes normal state, broad learning system...

While the existing reserving room before encryption (RRBE) based reversible data hiding in encrypted image (RDHEI) schemes can achieve decent embedding capacity, capacity of vacating by (VRBE) is relatively low. To address this issue, paper proposes a generalized framework for high-capacity RDHEI both RRBE and VRBE cases. First, an efficient generation algorithm (ERGA) designed to produce large using pixel prediction entropy encoding. Then, we propose two schemes, one RRBE, another VRBE. In...

The amount of multi-modal data available on the Internet is enormous. Cross-modal hash retrieval maps heterogeneous cross-modal into a single Hamming space to offer fast and flexible services. However, existing methods mainly rely feature-level similarity between ignore relationship relative rankings label-level fine-grained neighboring instances. To overcome these issues, we propose novel <underline xmlns:mml="http://www.w3.org/1998/Math/MathML"...

In strip steel production, detecting surface defects is crucial for ensuring product quality and optimizing production line efficiency. However, complicated by the variations in size, complex structures, wide range of defect morphologies present steel. To tackle these challenges, this paper proposes a detection network via adaptive focusing features (AFF-Net). Firstly, an feature block (AFF-Block) proposed, which applies "Diffusion-Aggregation" thought. This repositions adaptively assigns...

In this letter, we make the first attempt to explore usage of gradient direction conduct perceptual quality assessment screen content images (SCIs). Specifically, proposed approach extracts based on local information image magnitude, which not only preserves consistency in regions, but also demonstrates sensitivities distortions introduced SCI. A deviation-based pooling strategy is subsequently utilized generate corresponding index. Moreover, investigate and demonstrate complementary...

Person re-identification (Re-ID) aims to match person images captured from two non-overlapping cameras. In this paper, a deep hybrid similarity learning (DHSL) method for Re-ID based on convolution neural network (CNN) is proposed. our approach, light CNN feature pair the input image simultaneously extracted. Then, both elementwise absolute difference and multiplication of are calculated. Finally, function designed measure between pair, which realized by group weight coefficients project...

In this paper, a novel image quality assessment (IQA) model for the screen content images (SCIs) is proposed by using multi-scale difference of Gaussian (MDOG). Motivated observation that human visual system (HVS) sensitive to edges while details can be better explored in different scales, exploits MDOG effectively characterize edge information reference and distorted SCIs at two respectively. Then, degree similarity measured terms smaller-scale map. Finally, strength computed based on...

3D point clouds associated with attributes are considered as a promising paradigm for immersive communication. However, the corresponding compression schemes this media still in infant stage. Moreover, contrast to conventional image/video compression, it is more challenging task compress cloud data, arising from irregular structure. In paper, we propose novel and effective scheme of voxelized clouds. first stage, an input divided into blocks equal size. Then, deal structure clouds,...

A densely-sampled light field (LF) is highly desirable in various applications, such as 3-D reconstruction, post-capture refocusing and virtual reality. However, it costly to acquire data. Although many computational methods have been proposed reconstruct a LF from sparsely-sampled one, they still suffer either low reconstruction quality, efficiency, or the restriction on regularity of sampling pattern. To this end, we propose novel learning-based method, which accepts LFs with irregular...

This paper presents a new full-reference image quality assessment (IQA) method for conducting the perceptual evaluation of light field (LF) images, called symmetry and depth feature-based model (SDFM). Specifically, radial transform is first employed on luminance components reference distorted LF images to extract their features capturing spatial each view an image. Second, feature extraction scheme designed explore geometry information inherited in modeling its structural consistency across...

In this paper, a novel full-reference <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">image quality assessment</i> (IQA) method for evaluating the of distorted xmlns:xlink="http://www.w3.org/1999/xlink">light field</i> (LF) image against its reference LF is proposed, called xmlns:xlink="http://www.w3.org/1999/xlink">log-Gabor feature-based</i> light field coherence (LGF-LFC). Based on fact that to compare two images, it essentially boils down...

In general, vehicle images have varying resolutions due to vehicles' movements and different camera settings. However, most existing reidentification models are single-resolution deep networks trained with preuniformly resizing images, which underestimate adverse effects of lead unsatisfactory performance. A straightforward solution for dealing is train multiple models. Each model independently a specific resolution. this requires significant overhead ignores intrinsic associations among...

This paper studies a diffusion-based framework to address the low-light image enhancement problem. To harness capabilities of diffusion models, we delve into this intricate process and advocate for regularization its inherent ODE-trajectory. be specific, inspired by recent research that low curvature ODE-trajectory results in stable effective process, formulate term anchored intrinsic non-local structures data, i.e., global structure-aware regularization, which gradually facilitates...