Chinese Journal of Liquid Crystals and Displays, Volume. 37, Issue 10, 1334(2022)
Gradient-aware based single image super-resolution
Figures & Tables(7)
Fig. 3. Results compared visually with perceptual-driven SR methods（×4）
Table 1. Evaluation index of ablation experiment（Set14×4）
View tableView in ArticleTable 1. Evaluation index of ablation experiment（Set14×4）
Model name Grad-branch Non-concat Interim-conv Result RCAN_sub × × × LPIPS：0.133 2 PSNR：26.815 Model_1 √ × × LPIPS：0.132 0 PSNR：26.822 Model_2 √ √ × LPIPS：0.133 4 PSNR：26.844 GASR √ √ √ LPIPS：0.129 6（↓） PSNR：26.857（↑）
Table 2. Results compared with advanced perception-driven SR method（×4）
View tableView in ArticleTable 2. Results compared with advanced perception-driven SR method（×4）
Dataset Metric Bicubic SFTGAN SRGAN ESRGAN SPSR GASR（Ours） #Params. 1.8M 0.7M 16.7M 24.8M 4.3M FLOPs 0.8G 7.0G 89.7G 265.1G 26.6G Set5 PI 7.369 9 3.758 7 3.982 0 3.752 2 3.274 3 3.636 4 LPIPS 0.340 7 0.089 0 0.088 2 0.074 8 0.064 4 0.070 2 PSNR 28.420 29.932 29.168 30.454 30.400 30.517 SSIM 0.824 5 0.866 5 0.861 3 0.867 7 0.862 7 0.872 0 Set14 PI 7.026 8 2.906 3 3.085 1 2.926 1 2.903 6 3.027 2 LPIPS 0.439 3 0.148 1 0.166 3 0.132 9 0.131 8 0.129 6 PSNR 26.100 26.223 26.171 26.276 26.640 26.857 SSIM 0.785 0 0.785 4 0.784 1 0.778 3 0.793 0 0.798 1 BSD100 PI 7.002 6 2.377 4 2.545 9 2.479 3 2.351 0 2.427 3 LPIPS 0.524 9 0.176 9 0.198 0 0.161 4 0.161 1 0.161 6 PSNR 25.961 25.505 25.459 25.317 25.505 25.698 SSIM 0.667 5 0.654 9 0.648 5 0.650 6 0.657 6 0.664 2 General100 PI 7.936 5 4.287 8 4.375 7 4.323 4 4.099 1 4.250 7 LPIPS 0.352 8 0.103 0 0.105 5 0.087 9 0.086 3 0.085 1 PSNR 28.018 29.026 28.575 29.412 29.414 29.744 SSIM 0.828 2 0.850 8 0.854 1 0.854 6 0.853 7 0.861 4
Table 3. Comparison results on RealSR datasets（×4）
View tableView in ArticleTable 3. Comparison results on RealSR datasets（×4）
Method LPIPS PI PSNR SSIM Bibubic 0.485 3 8.189 0 27.507 0.865 9 ESRGAN 0.428 8 7.597 5 27.799 0.874 4 USRGAN 0.422 6 7.592 8 27.867 0.877 6 GASR（Ours） 0.417 9 7.503 7 27.805 0.876 4
Tools
Get Citation
Copy Citation Text
Le ZHOU, Long XU, Xiao-yan LIU, Xin-ze ZHANG, Xuan-de ZHANG. Gradient-aware based single image super-resolution[J]. Chinese Journal of Liquid Crystals and Displays, 2022, 37(10): 1334
Paper Information
Category: Research Articles
Received: Mar. 11, 2022
Accepted: --
Published Online: Oct. 10, 2022
The Author Email: Xuan-de ZHANG (zhangxuande@sust.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20