List your company
ResearchFormerly WizWand

Gaussian Denoising on BSD68 (Specific Noise Levels)

28.32PSNR (sigma=50)CEA-IR
20.977622.883824.7926.6962Mar 6, 2023Sep 15, 2023Mar 26, 2024Oct 5, 2024Apr 16, 2025Oct 26, 2025May 7, 2026
Updated 6mo ago

Evaluation results

MethodLinks
CEA-IRParams.=22M, Model Capacity=Heavy2026.0528.3234.2231.58
CEA-IR-SParams.=10M, Model Capacity=Light2026.0528.2734.1631.51
Res-HAIRParams.=29M, Model Capacity=Heavy2026.0528.2434.1631.51
VLU-NetParams.=35M, Model Capacity=Heavy2026.0528.2334.1331.48
LoRA-IRParams.=–, Model Capacity=Heavy2026.0528.1834.0631.42
MoCE-IRParams.=25M, Model Capacity=Heavy2026.0528.1834.1131.45
UniProcessorParams.=1002M, Model Capacity=Heavy2026.0528.1734.0831.42
MoCE-IR-SParams.=11M, Model Capacity=Light2026.0528.1634.0831.42
PromptIR+ArtParams.=33M, Model Capacity=Heavy2026.0528.1434.0631.42
GridformerParams.=34M, Model Capacity=Heavy2026.0528.1133.9331.37
PromptIRParams.=36M, Model Capacity=Heavy2026.0528.0633.9831.31
AirNetParams.=9M, Model Capacity=Light2026.052833.9231.26
MPRNetParams.=16M, Model Capacity=Heavy2026.0527.5633.5430.89
Restormer+Ours-cTraining protocol=Single model for all noise levels, Pre-trained CLIP model=true2024.0326.9132.1129.78
GRL-B+Ours-cTraining protocol=Separate model per noise level, Pre-trained CLIP model=true2024.0326.9132.2929.76
DLParams.=2M, Model Capacity=Light2026.0526.933.0530.41
Restormer+Ours-cTraining protocol=Separate model per noise level, Pre-trained CLIP model=true2024.0326.8632.2429.81
DRUNet+Ours-cTraining protocol=Single model for all noise levels, Pre-trained CLIP model=true2024.0326.8432.229.73
KBNet_sMACs=69G2023.0326.6531.9829.54
RestormerMACs=141G2023.0326.6231.9629.52
RestormerTraining protocol=Single model for all noise levels2024.0326.6231.9529.51
RestormerTraining protocol=Separate model per noise level2024.0326.6231.9629.52
GRL-BTraining protocol=Separate model per noise level2024.0326.63229.54
DRUNetMACs=144G2023.0326.5931.9129.48
DRUNetTraining protocol=Single model for all noise levels2024.0326.5931.9129.48
SwinIRMACs=759G2023.0326.5831.9729.5
DeamNetMACs=146G2023.0326.5431.9129.44
MWCNN2023.0326.5331.8629.41
DAGLMACs=256G2023.0326.5131.9329.46
FOCNet2023.0326.531.8329.38
RNANMACs=496G2023.0326.48——
NLRN2023.0326.4731.8829.41
FDGANParams.=–, Model Capacity=Light2026.0526.4330.2528.81
BRDNetParams.=–, Model Capacity=Light2026.0526.3432.2629.76
FFDNet2023.0326.2931.6329.19
DnCNNMACs=37G2023.0326.2331.6229.16
IRCNN2023.0326.1931.6329.15
DA-CLIPParams.=125M, Model Capacity=Heavy2026.0522.2930.0224.86
LPNetParams.=–, Model Capacity=Light2026.0521.2626.4724.77