A5092340527- Advanced Image Processing Techniques
- Advanced Vision and Imaging
- Video Coding and Compression Technologies
- Synthetic Aperture Radar (SAR) Applications and Techniques
- Domain Adaptation and Few-Shot Learning
- Advanced Image and Video Retrieval Techniques
- Remote-Sensing Image Classification
- Advanced Optical Imaging Technologies
- Advanced SAR Imaging Techniques
- Industrial Vision Systems and Defect Detection
- Geological and Geophysical Studies
- Advanced Neural Network Applications
Xi’an University of Posts and Telecommunications
2023
University Town of Shenzhen
2022-2023
Tsinghua University
2022-2023
Shanghai University of Electric Power
2023
PowerChina (China)
2023
Light field image becomes one of the most promising media types for immersive video applications. In this paper, we propose a novel end-to-end spatial-angular-decorrelated network (SADN) high-efficiency light compression. Different from existing methods that exploit either spatial or angular consistency in image, SADN decouples and information by dilation convolution stride spatial-angular interaction, performs feature fusion to compress jointly. To train stable robust algorithm, large-scale...
Focused plenoptic cameras can record spatial and angular information of the light field (LF) simultaneously with higher resolution relative to traditional cameras, which facilitate various applications in computer vision. However, existing image compression methods present ineffectiveness captured images due complex micro-textures generated by microlens relay imaging long-distance correlations among microimages. In this article, a lossy end-to-end learning architecture is proposed compress...
Object detection is a fundamental part of autonomous driving algorithms, and with the promotions transformers in couple years, numerous computer vision tasks are integrating into object detectors to acquire better generalization ability. Building pure transformer-based detector seems be wonderful choice; however, not omnipotent, they come painful drawbacks. Its operator, multi-head self-attention (MHSA), suffers from need for computational resources due its quadratic complexity, which...
The plenoptic 2.0 video can record a time-varying dense light field, which benefits many immersive visual applications such as AR/VR. However, traditional inter motion estimation methods perform inefficiently in kinds of sequences due to the distinctive temporal characteristics caused by imaging principle. In this paper, microimage-based two- step search (MTSS) is proposed achieve better trade-off between coding performance and complexity. Based on microimage focus variation analysis dynamic...
The object detection algorithms are the cornerstones of autonomous driving systems, they mostly based on convolutional neural networks (CNNs) with one or two stages.Since its strong correlation life safety driver, accuracy detectors is crucial and limited by foundation, CNN, which hard to improve nowadays.But at same time, basic transformer shows better performance compared advanced CNN.To accuracy, using transformers seems be a choice.However, most transformer-based only backbone...