Laser & Optoelectronics Progress, Volume. 60, Issue 4, 0411002(2023)
Image Super-Resolution Reconstruction Algorithm Based on Enhanced Multi-Scale Residual Network
Figures & Tables(10)
Fig. 8. Comparison of network parameters and PSNR correspondence for different algorithms
Table 1. Comparison between EMSRN-NOCA and EMSRN
View tableTable 1. Comparison between EMSRN-NOCA and EMSRN
Algorithm Scale Manga109 PSNR/SSIM EMSRN-NOCA ×2 39.07/0.9777 EMSRN ×2 39.20/0.9782 EMSRN-NOCA ×3 34.01/0.9478 EMSRN ×3 34.16/0.9483 EMSRN-NOCA ×4 30.85/0.9142 EMSRN ×4 30.95/0.9148 EMSRN-NOCA ×8 24.74/0.7852 EMSRN ×8 24.72/0.7857
Table 2. PSNR and SSIM values of different algorithms
View tableTable 2. PSNR and SSIM values of different algorithms
Algorithm Scale Set5
PSNR/SSIM
Set14
PSNR/SSIM
BSD100
PSNR/SSIM
Urban100
PSNR/SSIM
Manga109
PSNR/SSIM
Bicubic ×2 33.69/0.9284 30.34/0.8675 29.57/0.8434 26.88/0.8438 30.82/0.9332 SRCNN ×2 36.31/0.9535 32.26/0.9053 31.16/0.8859 29.30/0.8939 35.16/0.9663 FSRCNN ×2 36.78/0.9561 32.57/0.9089 31.38/0.8894 29.74/0.9009 36.26/0.9700 ESPCN ×2 36.47/0.9544 32.32/0.9067 31.17/0.8867 29.21/0.8925 35.28/0.9666 VDSR ×2 37.16/0.9582 32.87/0.9126 31.75/0.8951 30.74/0.9146 36.42/0.9730 LapSRN ×2 36.91/0.9573 32.71/0.9105 31.59/0.8922 30.26/0.9090 36.09/0.9717 EDSR ×2 38.11/0.9601 33.92 / 0.919532.32 / 0.9013-/- -/- MSRN ×2 38.09/0.9607 33.73/0.9182 32.22/0.9001 32.29/0.9300 38.65/0.9771 IMDN ×2 37.91/0.9594 33.59/0.9169 32.15/0.8987 32.14/0.9274 38.79/0.9764 CFSRCNN ×2 37.79/0.9591 33.50/0.9165 32.10/0.8987 32.07/0.9273 38.16/0.9751 PAN ×2 38.00/0.9605 33.59/0.9181 32.18/0.8997 32.01/0.9273 38.70/0.9773 EMSRN（proposed） ×2 38.19/0.9613 33.78/0.9195 32.30/0.9012 32.74 / 0.934239.20 / 0.9782Bicubic ×3 30.41/0.8655 27.64/0.7722 27.21/0.7344 24.46/0.7411 26.96/0.8555 SRCNN ×3 32.60/0.9088 29.21/0.8198 28.30/0.7840 26.04/0.7955 30.09/0.9098 FSRCNN ×3 32.51/0.9054 29.17/0.8181 28.24/0.7821 25.97/0.7917 30.00/0.9051 ESPCN ×3 32.56/0.9073 29.19/0.8195 28.26/0.7834 25.98/0.7929 30.01/0.9063 VDSR ×3 33.54/0.9214 29.69/0.8315 28.73/0.7963 27.05/0.8265 31.55/0.9312 EDSR ×3 34.65/0.9282 30.52/0.8462 29.25 / 0.8093-/- -/- MSRN ×3 34.47/0.9276 30.38/0.8434 29.13/0.8062 28.31/0.8558 33.58/0.9452 IMDN ×3 34.32/0.9259 30.31/0.8409 29.07/0.8036 28.15/0.8510 33.58/0.9434 CFSRCNN ×3 34.23/0.9256 30.26/0.8409 29.02/0.8033 28.03/0.8495 33.27/0.9418 PAN ×3 34.40/0.9272 30.36/0.8422 29.10/0.8049 28.11/0.8510 33.58/0.9448 EMSRN（proposed） ×3 34.65 / 0.929030.54/0.8461 29.24/0.8089 28.70 / 0.863434.16 / 0.9483Bicubic ×4 28.43/0.8022 26.10/0.6936 25.97/0.6517 23.14/0.6599 24.91/0.7826 SRCNN ×4 30.22/0.8597 27.40/0.7489 26.78/0.7074 24.29/0.7141 27.10/0.8457 FSRCNN ×4 30.44/0.8595 27.51/0.7507 26.85/0.7090 24.44/0.7188 27.42/0.8432 ESPCN ×4 30.25/0.8566 27.37/0.7487 26.77/0.7072 24.26/0.7114 27.00/0.8398 VDSR ×4 31.18/0.8819 28.00/0.7673 27.19/0.7229 25.09/0.7494 28.50/0.8812 LapSRN ×4 31.52/0.8879 28.17/0.7729 27.28/0.7278 25.26/0.7596 28.85/0.8900 EDSR ×4 32.46/0.8968 28.80 / 0.787627.71 / 0.7420-/- -/- MSRN ×4 32.19/0.8952 28.63/0.7837 27.61/0.7377 26.16/0.7894 30.53/0.9093 IMDN ×4 32.19/0.8936 28.57/0.7803 27.54/0.7342 26.03/0.7829 30.44/0.9065 CFSRCNN ×4 32.06/0.8920 28.57/0.7801 27.52/0.7331 26.02/0.7823 30.30/0.9048 PAN ×4 32.13/0.8948 28.61/0.7822 27.59/0.7363 26.11/0.7854 30.51/0.9095 EMSRN（proposed） ×4 32.43/0.8980 28.77/0.7863 27.69/0.7408 26.51 / 0.799530.95 / 0.9148Bicubic ×8 24.40/0.6045 23.19/0.5110 23.67/0.4808 20.74/0.4841 21.46/0.6138 SRCNN ×8 24.10/0.6599 23.02/0.5704 23.53/0.5479 20.68/0.5157 21.41/0.6470 FSRCNN ×8 25.20/0.6832 23.77/0.5879 24.08/0.5599 21.21/0.5342 22.21/0.6630 ESPCN ×8 25.41/0.6895 23.92/0.5923 24.13/0.5629 21.31/0.5382 22.32/0.6677 VDSR ×8 25.79/0.7183 24.20/0.6084 24.37/0.5744 21.58/0.5614 22.82/0.7056 LapSRN ×8 25.96/0.7276 24.35/0.6168 24.49/0.5811 21.74/0.5727 23.12/0.7202 MSRN ×8 26.87/0.7714 24.83/0.6381 24.74/0.5950 22.32/0.6122 24.39/0.7743 EMSRN（proposed） ×8 26.88 / 0.775924.99 / 0.643424.79 / 0.598722.55 / 0.623024.72 / 0.7857
Tools
Get Citation
Copy Citation Text
Jiao Xu, Sannan Yuan. Image Super-Resolution Reconstruction Algorithm Based on Enhanced Multi-Scale Residual Network[J]. Laser & Optoelectronics Progress, 2023, 60(4): 0411002
Paper Information
Category: Imaging Systems
Received: Nov. 5, 2021
Accepted: Dec. 21, 2021
Published Online: Feb. 14, 2023
The Author Email: Xu Jiao (xj15240039674@163.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20