Fudi Yi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Fudi Yi has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- He Jiang
School of Mines, China University of Mining and Technology, Xuzhou 221116, China
4 shared publications - Hao Gu
China University of Mining and Technology
3 shared publications - Wen Cui
School of Information and Control Engineering, China University of Mining and Technology, Xuzhou 221116, China
3 shared publications - Zhou Zheng
China University of Mining and Technology
3 shared publications - Baihui Deng
China University of Mining and Technology
1 shared publication - Bingyuan Su
China University of Mining and Technology
1 shared publication
Latest Publications
- 1 May 2026
- Published by Springer Nature
- p. 1-8
TaylorNet: Single Image Detail Enhancement Based on Taylor Series Decomposition
- 1 May 2026
- Published by Springer Nature
- p. 1-8
TaylorNet: Single Image Detail Enhancement Based on Taylor Series Decomposition
- 1 May 2026
- Published by Springer Nature
- p. 304-311
QA-TransUNet: Paying More Attention to the Shape of Semantic Objects
- 1 May 2026
- Published by Springer Nature
- p. 304-311
QA-TransUNet: Paying More Attention to the Shape of Semantic Objects
- 24 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (7) , 14051-14066
Tree-Shaped Recursive Network: Rethinking Directional Features in Lightweight Single Image Super Resolution
Single-image super-resolution (SISR) is a post-processing technology for the Internet of Things (IoT) devices that reduces bandwidth usage and improves communication efficiency. While most SISR models perform similarly on images with simple textures, their differences become more pronounced when handling complex directional textures, which are challenging to reconstruct. In view of this, we propose a tree-shaped recursive network (TRN) for SISR that utilizes a tree-shaped topology and a non-local coordinate attention (NLCA) mechanism to capture global directional textures. First, a binary tree residual block (BTRB) is introduced to preserve the attention and residual features without increasing the number of parameters. Second, an NLCA mechanism is applied to integrate non-local and directional features to effectively capture global information while reducing computational complexity. Third, a directional attention fusion technique that integrates attention and residual features is presented. Extensive experimental results show that TRN outperforms other models in terms of evaluation metrics such as peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). Notably, on the $\times 2$ Urban100 dataset, the PSNR of TRN is 0.24 dB higher than that of the popular model CFIN, which highlights its potential and usefulness.
- 24 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (7) , 14051-14066
Tree-Shaped Recursive Network: Rethinking Directional Features in Lightweight Single Image Super Resolution
Single-image super-resolution (SISR) is a post-processing technology for the Internet of Things (IoT) devices that reduces bandwidth usage and improves communication efficiency. While most SISR models perform similarly on images with simple textures, their differences become more pronounced when handling complex directional textures, which are challenging to reconstruct. In view of this, we propose a tree-shaped recursive network (TRN) for SISR that utilizes a tree-shaped topology and a non-local coordinate attention (NLCA) mechanism to capture global directional textures. First, a binary tree residual block (BTRB) is introduced to preserve the attention and residual features without increasing the number of parameters. Second, an NLCA mechanism is applied to integrate non-local and directional features to effectively capture global information while reducing computational complexity. Third, a directional attention fusion technique that integrates attention and residual features is presented. Extensive experimental results show that TRN outperforms other models in terms of evaluation metrics such as peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). Notably, on the $\times 2$ Urban100 dataset, the PSNR of TRN is 0.24 dB higher than that of the popular model CFIN, which highlights its potential and usefulness.
- Open Access
Tracking-Based Denoising: A Trilateral Filter-Based Denoiser for Real-World Surveillance Video in Extreme Low-Light Conditions
Video denoising in extremely low-light surveillance scenarios is a challenging task in computer vision, as it suffers from harsh noise and insufficient signal to reconstruct fine details. The denoising algorithm for these scenarios encounters challenges such as the lack of ground truth, and the noise distribution in the real world is far more complex than in a normal scene. Consequently, recent state-of-the-art (SOTA) methods like VRT and Turtle for video denoising perform poorly in this low-light environment. Additionally, some methods rely on raw video data, which is difficult to obtain from surveillance systems. In this paper, a denoising method is proposed based on the trilateral filter, which aims to denoise real-world low-light surveillance videos. Our trilateral filter is a weighted filter, allocating reasonable weights to different inputs to produce an appropriate output. Our idea is inspired by an experimental finding: noise on stationary objects can be easily suppressed by averaging adjacent frames. This led us to believe that if we can track moving objects accurately and filter along their trajectories, the noise may be effectively removed. Our proposed method involves four main steps. First, coarse motion vectors are obtained by bilateral search. Second, an amplitude-phase filter is used to judge and correct erroneous vectors. Third, these vectors are refined by a full search in a small area for greater accuracy. Finally, the trilateral filter is applied along the trajectory to denoise the noisy frame. Extensive experiments have demonstrated that our method achieves superior performance in terms of visual effects and quantitative tests.
Open AccessTracking-Based Denoising: A Trilateral Filter-Based Denoiser for Real-World Surveillance Video in Extreme Low-Light Conditions
Video denoising in extremely low-light surveillance scenarios is a challenging task in computer vision, as it suffers from harsh noise and insufficient signal to reconstruct fine details. The denoising algorithm for these scenarios encounters challenges such as the lack of ground truth, and the noise distribution in the real world is far more complex than in a normal scene. Consequently, recent state-of-the-art (SOTA) methods like VRT and Turtle for video denoising perform poorly in this low-light environment. Additionally, some methods rely on raw video data, which is difficult to obtain from surveillance systems. In this paper, a denoising method is proposed based on the trilateral filter, which aims to denoise real-world low-light surveillance videos. Our trilateral filter is a weighted filter, allocating reasonable weights to different inputs to produce an appropriate output. Our idea is inspired by an experimental finding: noise on stationary objects can be easily suppressed by averaging adjacent frames. This led us to believe that if we can track moving objects accurately and filter along their trajectories, the noise may be effectively removed. Our proposed method involves four main steps. First, coarse motion vectors are obtained by bilateral search. Second, an amplitude-phase filter is used to judge and correct erroneous vectors. Third, these vectors are refined by a full search in a small area for greater accuracy. Finally, the trilateral filter is applied along the trajectory to denoise the noisy frame. Extensive experiments have demonstrated that our method achieves superior performance in terms of visual effects and quantitative tests.