Gaussian Denoising on Set12 (Sigma Noise Levels 15, 25, 50)
28.37PSNR (sigma=50)GRL-B+Ours-c
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GRL-B+Ours-cTraining protocol=Separate model per noise level, Pre-trained CLIP model=true2024.03 | 28.37 | 33.74 | 31.3 | |
| Restormer+Ours-cTraining protocol=Single model for all noise levels, Pre-trained CLIP model=true2024.03 | 28.36 | 33.57 | 31.28 | |
| Restormer+Ours-cTraining protocol=Separate model per noise level, Pre-trained CLIP model=true2024.03 | 28.35 | 33.7 | 31.29 | |
| DRUNet+Ours-cTraining protocol=Single model for all noise levels, Pre-trained CLIP model=true2024.03 | 28.27 | 33.51 | 31.18 | |
| KBNet_sMACs=69G2023.03 | 28.04 | 33.4 | 31.08 | |
| GRL-BTraining protocol=Separate model per noise level2024.03 | 28.03 | 33.47 | 31.12 | |
| RestormerTraining protocol=Single model for all noise levels2024.03 | 28.01 | 33.35 | 31.04 | |
| RestormerMACs=141G2023.03 | 28 | 33.42 | 31.08 | |
| RestormerTraining protocol=Separate model per noise level2024.03 | 28 | 33.42 | 31.08 | |
| SwinIRMACs=759G2023.03 | 27.91 | 33.36 | 31.01 | |
| DRUNetMACs=144G2023.03 | 27.9 | 33.25 | 30.94 | |
| DRUNetTraining protocol=Single model for all noise levels2024.03 | 27.9 | 33.25 | 30.94 | |
| DAGLMACs=256G2023.03 | 27.81 | 33.28 | 30.93 | |
| MWCNN2023.03 | 27.74 | 33.15 | 30.79 | |
| DeamNetMACs=146G2023.03 | 27.74 | 33.19 | 30.81 | |
| RNANMACs=496G2023.03 | 27.7 | — | — | |
| FOCNet2023.03 | 27.68 | 33.07 | 30.73 | |
| NLRN2023.03 | 27.64 | 33.16 | 30.8 | |
| FFDNet2023.03 | 27.32 | 32.75 | 30.43 | |
| DnCNNMACs=37G2023.03 | 27.18 | 32.67 | 30.35 | |
| IRCNN2023.03 | 27.12 | 32.76 | 30.37 |