一种超分辨率遥感数据重构方法
By employing a multi-step generalization framework and an attention-based image correction network and fuzzy kernel estimation network, the problem of weak generalization ability of remote sensing images under unknown fuzzy kernel degradation is solved, achieving efficient super-resolution reconstruction of remote sensing images and improving the model's adaptability and accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2023-03-30
- Publication Date
- 2026-07-17
AI Technical Summary
Existing deep learning-based super-resolution methods for remote sensing images have weak generalization ability when faced with unknown blur kernel degradation, resulting in performance degradation and difficulty in effectively handling complex degradation phenomena in remote sensing images.
A multi-step generalization framework is designed, including an image correction network, a fuzzy kernel estimation network, and a pre-trained super-resolution network. The fuzzy kernel estimation and image correction are gradually optimized through a multi-step iterative process. The network performance is optimized by utilizing the dual-path fusion module of the attention mechanism and the prior information of the fuzzy kernel, combined with a multi-step loss function.
It improves the performance of blind super-resolution reconstruction of remote sensing images, reduces the generalization difficulty of unknown blur kernel degradation, and improves the accuracy and adaptability of the model, enabling more accurate reconstruction of high-resolution remote sensing images.
Smart Images
Figure CN116385264B_ABST