A5086371835- Video Coding and Compression Technologies
- Image and Video Quality Assessment
- Advanced Image Processing Techniques
- Advanced Vision and Imaging
- Advanced Data Compression Techniques
- Visual Attention and Saliency Detection
- Image Enhancement Techniques
- Visual perception and processing mechanisms
- Advanced Neural Network Applications
- Music and Audio Processing
- Advanced Image Fusion Techniques
- Color Science and Applications
- Advanced Image and Video Retrieval Techniques
- Image and Signal Denoising Methods
- Video Surveillance and Tracking Methods
- Image Processing Techniques and Applications
- Music Technology and Sound Studies
- Video Analysis and Summarization
- Speech and Audio Processing
- Infrared Target Detection Methodologies
- 3D Surveying and Cultural Heritage
- Advanced Computational Techniques and Applications
- Phase Change Materials Research
- IoT Networks and Protocols
- Optical measurement and interference techniques
Hangzhou Dianzi University
2021-2024
Lishui University
2022-2024
Huazhong University of Science and Technology
2018-2022
Tianjin University of Commerce
2022
China Jiliang University
2016
Salient object detection of surface defects is one the defect tasks, which aims at highlighting regions from strip steel, magnetic tale, road, and so on. However, performance existing methods degrades dramatically when dealing with complex scenarios, such as low contrast various shapes. Therefore, in this article, we propose a novel saliency model, namely, localizing, focus, refinement network (LFRNet), consists semantic-guided localizing module, context-driven focus edge-aware (ER) module....
Change captioning aims to describe the differences between two similar images using natural language, significantly aiding in understanding and monitoring changes. This challenging task requires a fine-grained of subtle changes while resisting disturbances like viewpoint shifts illumination variations. Existing methods often rely solely on global difference features lack comprehensive alignment linguistic visual information, leading overlooking details generating semantic hallucinated...
The human visual system (HVS) is a hierarchical system, in which signals are processed hierarchically. In this paper, the HVS modeled as three-level communication and perception divided into three stages according to predictive coding theory. Then, novel just noticeable distortion (JND) estimation scheme proposed. perception, input predicted constantly spontaneously each hierarchy, neural response evoked by central residue inhibited surrounding residues. These two types' residues regarded...
In visual perception, the neural response is evoked by central stimulus and inhibited surrounding stimuli. These two stimuli can be regarded as positive negative incentives for perception. this letter, aftereffect of incorporated into just noticeable distortion (JND) estimation to avoid threshold overestimation issue. neuroscience, how an incentive affects observer measured surprise which estimated residual self-information according effective coding theory. Considering competition biases,...
Due to the limited storage capacity and network bandwidth, an efficient rate control (RC) algorithm becomes more critical for multi-view video coding (MVC). Based on inter-view dependency spatio-temporal correlation, a novel bit allocation method texture is proposed in this article. Firstly, considering that distortion base view (BV) directly transmitted dependent (DV) by skip mode, joint RD model built based dependency. model, precise power derived represent target bitrates relationship...
The just noticeable difference (JND) reveals the key characteristic of visual perception, which has been widely used in many perception-based image and video applications. Nevertheless, modulatory mechanism human system (HVS) not fully exploited JND threshold estimation, results existing models being accurate enough. In this article, by analyzing feedforward feedback behaviors HVS, an enhanced foveated (FJND) estimation model is proposed considering effects masking perception. contributions...
3D video coding extension of High Efficiency Video Coding (3D-HEVC) adopts many high complexity approaches to improve the performance depth video, which leads heavy computation. To alleviate computation burden, a fast intra scheme for map is proposed based on texture feature and spatial correlation. The unit (CU) sizes prediction modes are selected differently due different features. CU block divided into smooth block, edge by using gray-level co-occurrence matrix (GLCM) sobel operator....
The aftereffect of visual attention (VA) and foveated masking (FM) are two important HVS characteristics not fully exploited in conventional JND (FJND) models. This paper presents a fixation point estimation method to build adaptive weighting model, by which an improved FJND profile is developed with VA FM effects considered. proposed overcomes limitations the profiles. On one hand, profiles can't get accurate positions points. proposes prediction algorithm based on distribution cones...
Recently, deep learning technology has made significant progresses in high efficiency video coding (HEVC), especially in-loop filter. In this paper, we propose a dense inception attention network (DIA_Net) to delve into image information and model capacity. The DIA_Net contains multiple blocks which have different size kernels so as dig out various scales information. Meanwhile, mechanism including spatial channel is utilized fully exploit feature Further adopt residual structure deepen the...
Abstract Screen content video (SCV) consists primarily of text areas, computer graphics and other computer‐generated possesses unique perceptual characteristics. To compress SCVs more effectively with less reduction in subjective quality, characteristics are analyzed a redundancy (PR) model for SCV compression is proposed, including spatial PR (SPR), temporal (TPR) foveated (FPR) model. In SPR modeling, the divided into sharp edge (SE) areas non‐SE then estimated separately. TPR both...
In the latest video coding standard, Versatile Video Coding (H.266/VVC), a new quadtree with nested multi-type tree (QTMTT) block structure is proposed. QTMTT significantly improves performance, but more complex partitioning brings greater computational burden. To solve this problem, fast intra partition pattern pruning algorithm proposed using gray level co-occurrence matrix (GLCM) to calculate texture direction information of units, terminating horizontal or vertical split binary and...
The complexity of 3D-HEVC is fairly high due to quad tree structure and traversal searching in depth intra coding. In order reduce caused by coding unit (CU) size decision rate distortion optimization (RDO) process, a fast algorithm based on adaptive QP convolutional neural network (AQ-CNN) proposed this paper. For each CU, the automatically extracts deep feature information terminate CU partition early. Specially, AQ-CNN suitable for different QPs because has great influence connected into...