Gaussian Image Denoising on McMaster 2011 (test)
30.5PSNR (sigma=50)DMID-d
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DMID-dTraining strategy=Noise-blind (various noise levels)2023.07 | 30.5 | 35.72 | 33.49 | |
| ART*MACs (G)=4220, Inf. (ms)=OOM, Image Size=512 x 512 x 32024.01 | 30.31 | 35.68 | 33.41 | |
| RestormerMACs (G)=564, Inf. (ms)=350, Image Size=512 x 512 x 32024.01 | 30.3 | 35.61 | 33.34 | |
| RestormerTraining strategy=Noise-specific (separate models)2023.07 | 30.3 | 35.61 | 33.34 | |
| RestormerTraining strategy=Noise-blind (various noise levels)2023.07 | 30.29 | 35.55 | 33.31 | |
| SwinIRMACs (G)=2991, Inf. (ms)=1850, Image Size=512 x 512 x 32024.01 | 30.22 | 35.61 | 33.2 | |
| CGNetMACs (G)=444, Inf. (ms)=215, Image Size=512 x 512 x 32024.01 | 30.22 | 35.58 | 33.28 | |
| SwinIRTraining strategy=Noise-specific (separate models)2023.07 | 30.22 | 35.61 | 33.2 | |
| GRL-SMACs (G)=975, Inf. (ms)=680, Image Size=512 x 512 x 32024.01 | 30.18 | 35.32 | 33.29 | |
| DRUNetTraining strategy=Noise-blind (various noise levels)2023.07 | 30.08 | 35.4 | 33.14 | |
| CODEMACs (G)=180, Inf. (ms)=600, Image Size=512 x 512 x 32024.01 | 30.03 | 35.38 | 33.11 | |
| CODETraining strategy=Noise-specific (separate models)2023.07 | 30.03 | 35.38 | 33.11 | |
| IPTTraining strategy=Noise-specific (separate models)2023.07 | 29.98 | — | — | |
| RNANTraining strategy=Noise-specific (separate models)2023.07 | 29.72 | — | — | |
| BRDNetTraining strategy=Noise-specific (separate models)2023.07 | 29.52 | 35.08 | 32.75 | |
| RDNTraining strategy=Noise-specific (separate models)2023.07 | 29.38 | — | — | |
| DSNetTraining strategy=Noise-blind (various noise levels)2023.07 | 29.28 | 34.67 | 32.4 | |
| FFDNetTraining strategy=Noise-blind (various noise levels)2023.07 | 29.18 | 34.66 | 32.35 | |
| DnCNNTraining strategy=Noise-blind (various noise levels)2023.07 | 28.62 | 33.45 | 31.52 | |
| CBM3DTraining strategy=Noise-blind (various noise levels)2023.07 | 28.51 | 34.06 | 31.66 |