A5075321916- Advanced Vision and Imaging
- Advanced Image Processing Techniques
- Image Processing Techniques and Applications
- Image and Signal Denoising Methods
- Image Enhancement Techniques
- Medical Image Segmentation Techniques
- Infrared Target Detection Methodologies
- Satellite Image Processing and Photogrammetry
- Blind Source Separation Techniques
Shenzhen Academy of Robotics
2024
Shenzhen Institutes of Advanced Technology
2024
Chinese Academy of Sciences
2024
University of Macau
2020-2021
Disparity estimation is a popular topic in computer vision and has drawn increasing attention recent years. In this article, we propose new multi-stage network for the purpose of two to three-dimensional video conversion that contains training stages: an initial disparity as first stage depth-image-based rendering (DIBR) extra component form second stage. stage, revised end-to-end feature pyramid stereo network, which original non-pyramid structure replaced by bottom-up convolutional neural...
In recent years, numerous real-time stereo matching methods have been introduced, but they often lack accuracy. These attempt to improve accuracy by introducing new modules or integrating traditional methods. However, the improvements are only modest. this paper, we propose a novel strategy incorporating knowledge distillation and model pruning overcome inherent trade-off between speed As result, obtained that maintains performance while delivering high on edge devices. Our proposed method...
The soft-argmax operation is widely adopted in neural network-based stereo matching methods to enable differentiable regression of disparity. However, network trained with prone being multimodal due absence explicit constraint the shape probability distribution. Previous leverages Laplacian distribution and cross-entropy for training but failed effectively improve accuracy even compromises efficiency network. In this paper, we conduct a detailed analysis previous distribution-based propose...