一种退化核无关的渐进式解耦盲超分辨率方法及系统
By employing a progressively decoupled blind super-resolution method independent of degradation kernels, this method utilizes feature separation and a re-degradation network to optimize the decoupling between super-resolution features and degradation features. This addresses the problem of inaccurate degradation kernels in existing methods and improves image restoration performance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- WUHAN UNIV
- Filing Date
- 2023-12-08
- Publication Date
- 2026-07-17
AI Technical Summary
Existing blind super-resolution methods perform poorly when dealing with real images with complex degradation types. The degradation kernel estimated by the network is inaccurate and cannot make full use of low-resolution image information, which affects the prediction of high-resolution images.
A progressively decoupled blind super-resolution method independent of degradation kernel is adopted. Through a shallow feature extraction module, feature separation, parallel branch decoupling, and feature re-degradation network, the decoupling of super-resolution features and degradation features is gradually optimized to generate high-level super-resolution images.
It effectively reduces the error caused by inaccurate degradation kernels, improves the restoration of image details and textures, and enhances the performance of blind super-resolution algorithms.
Smart Images
Figure CN117788289B_ABST