Laser & Optoelectronics Progress, Volume. 60, Issue 14, 1410013(2023)
Super-Resolution Reconstruction of FY-4 Images Based on Matching Extraction and Cross-Scale Feature Fusion Network
Fig. 1. Sample examples of training set. (a) 2× network training set samples; (b) 4× network training set samples
Fig. 2. Matching and extraction module
Fig. 3. Image feature visualization. (a) LR images; (b) reference images
Fig. 4. Cross-scale module structure
Fig. 5. Structure of 2× network
Fig. 6. Basic module structure. (a) Encoder structure; (b) decoder structure
Fig. 7. Calculation quantity and parameter quantity of each method
Fig. 8. Spectral curve comparison. (a) Original 2000 m band statistical results; (b) statistical results after super-resolution; (c) histograms of 4, 5, 6 bands of original 2000 m data; (d) statistical results after super-resolution
Fig. 9. 2× super-resolution reconstruction results of each algorithm
Fig. 10. 4× super-resolution reconstruction results of each algorithm
|
|
Get Citation
Copy Citation Text
Zhengsong Lu, Xi Kan, Yan Li, Naiyuan Chen. Super-Resolution Reconstruction of FY-4 Images Based on Matching Extraction and Cross-Scale Feature Fusion Network[J]. Laser & Optoelectronics Progress, 2023, 60(14): 1410013
Category: Image Processing
Received: Jul. 6, 2022
Accepted: Sep. 13, 2022
Published Online: Jul. 17, 2023
The Author Email: Kan Xi (kanxi@nuist.edu.cn)