Table 6 Comparison of different encoder–decoder combinations on model performance and complexity on the Flare7K++ real test dataset. Best results are highlighted in bold.
Models | PSNR\(\uparrow\) | SSIM\(\uparrow\) | LPIPS\(\downarrow\) | G-PSNR\(\uparrow\) | S-PSNR\(\uparrow\) | Params (M) | MACs (G) |
|---|---|---|---|---|---|---|---|
SMEBlock NAFBlock | 27.833 | 0.896 | 0.0438 | 24.279 | 22.345 | 7.777 | 95.885 |
NAFBlock SMDBlock | 27.691 | 0.895 | 0.0456 | 24.028 | 22.516 | 7.081 | 95.676 |
All SMDBlock | 27.577 | 0.896 | 0.0436 | 23.997 | 22.398 | 6.981 | 110.963 |
All SMEBlock | 27.835 | 0.896 | 0.0435 | 24.370 | 22.740 | 8.420 | 98.826 |
All NAFBlock | 27.317 | 0.893 | 0.0464 | 24.027 | 21.891 | 6.439 | 92.735 |
SMFR-Net (ours) | 27.925 | 0.901 | 0.0407 | 24.204 | 22.894 | 7.981 | 103.888 |
