针对多退化类别恢复的盲超分方法与装置
By decomposing the image restoration process into three modules—noise feature suppression, texture feature enhancement, and sampling restoration—and compensating for different degradation types, this approach addresses the performance limitations of existing blind super-resolution algorithms in multi-degradation scenarios, achieving more efficient image restoration and texture detail enhancement.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- FIBERHOME TELECOMMUNICATION TECHNOLOGIES CO LTD
- Filing Date
- 2023-07-06
- Publication Date
- 2026-07-17
AI Technical Summary
Existing blind super-resolution algorithms fail to specifically suppress the effects of different degradation types when dealing with various degradation types, leading to increased recovery difficulty and decreased performance, especially when dealing with image information loss caused by noise and blur kernels.
The image restoration process is decomposed into three modules: a noise feature suppression module, a texture feature enhancement module, and a sampling restoration module. The modules compensate for downsampling, blur kernel, and noise, respectively, and improve texture detail restoration through gradient weighted loss. A noise feature suppression module based on an attention mechanism and a densely connected texture feature enhancement module are designed.
It effectively improves the performance of the super-resolution model, enhances its robustness to image degradation processes in the real world, and improves the efficiency of image restoration and the reconstruction effect of texture details.
Smart Images
Figure CN116739904B_ABST