Image Denoising on DND
40.15PSNRUnfoldLDM
Evaluation results
| Method | Links | ||
|---|---|---|---|
| UnfoldLDMSources=Ours2025.11 | 40.15 | 0.964 | |
| CondformerMethod Category=DNN models (Supervised)2025.12 | 40.1 | 0.956 | |
| MambaIR2024.02 | 40.04 | 0.956 | |
| MambaIRSources=ECCV242025.11 | 40.04 | 0.956 | |
| RestormerTraining Supervision=Supervised (Real pairs)2023.03 | 40.03 | 0.956 | |
| Restormer2024.02 | 40.03 | 0.956 | |
| RestormerMethod Category=DNN models (Supervised)2025.12 | 40.03 | 0.956 | |
| RestormerLearning Paradigm=Supervised2026.04 | 40.03 | 0.956 | |
| Uformer2024.02 | 39.96 | 0.956 | |
| DeepSNMethod Category=DNN models (Supervised)2025.12 | 39.92 | 0.956 | |
| DeepSN-NetSources=TPAMI252025.11 | 39.92 | 0.956 | |
| MIRNet2022.03 | 39.88 | 0.956 | |
| ADFNetSources=AAAI232025.11 | 39.87 | 0.955 | |
| VIRNetSources=TPAMI242025.11 | 39.83 | 0.954 | |
| MPRNet2022.03 | 39.8 | 0.954 | |
| MPRNet2024.02 | 39.8 | 0.954 | |
| DAGL2022.03 | 39.77 | 0.956 | |
| DAGL2024.02 | 39.77 | 0.956 | |
| DIDNMethod Category=DNN models (Supervised)2025.12 | 39.64 | 0.953 | |
| DeamNetAdditional training data=true2022.03 | 39.63 | 0.953 | |
| DeamNet2024.02 | 39.63 | 0.953 | |
| DeamNetType=Supervised, Train Data=Paired Noisy-Clean, #Param=2.23M2025.12 | 39.63 | 0.9531 | |
| SADNetAdditional training data=true2022.03 | 39.59 | 0.952 | |
| DANet+Additional training data=true2022.03 | 39.58 | 0.955 | |
| ClipDeMethod Category=DNN models (Supervised)2025.12 | 39.57 | 0.954 | |
| CycleISPAdditional training data=true2022.03 | 39.56 | 0.956 | |
| SRMNet2022.03 | 39.44 | 0.951 | |
| TM-BSN (D)Learning Paradigm=Self-Supervised, Evaluation Protocol=Fully self-supervised, Model Variant=D2026.04 | 39.41 | 0.949 | |
| VDN2022.03 | 39.39 | 0.952 | |
| VDNTraining Supervision=Supervised (Real pairs)2023.03 | 39.38 | 0.952 | |
| VDN2019.08 | 39.38 | 0.9518 | |
| VDNetType=Supervised, Train Data=Paired Noisy-Clean, #Param=7.82M2025.12 | 39.38 | 0.9518 | |
| VDNLearning Paradigm=Supervised2026.04 | 39.38 | 0.953 | |
| AINDNetAdditional training data=true2022.03 | 39.37 | 0.951 | |
| Baseline, N2CTraining Supervision=Supervised (Real pairs)2023.03 | 39.37 | 0.954 | |
| RIDNetAdditional training data=true2022.03 | 39.26 | 0.953 | |
| RIDNetType=Supervised, Train Data=Paired Noisy-Clean, #Param=1.50M2025.12 | 39.25 | 0.9528 | |
| TBSNLearning Paradigm=Self-Supervised, Evaluation Protocol=Fully self-supervised2026.04 | 39.08 | 0.945 | |
| TM-BSN (D)Learning Paradigm=Self-Supervised, Evaluation Protocol=Standard, Model Variant=D2026.04 | 38.96 | 0.947 | |
| NLHCCMethod Category=Traditional denoisers2025.12 | 38.85 | 0.953 | |
| PUCALearning Paradigm=Self-Supervised, Evaluation Protocol=Fully self-supervised2026.04 | 38.83 | 0.942 | |
| APR (RD)Learning Paradigm=Self-Supervised, Evaluation Protocol=Fully self-supervised, Model Variant=(RD)2026.04 | 38.83 | 0.944 | |
| DnCNNTraining Supervision=Supervised (Real pairs)2023.03 | 38.73 | 0.945 | |
| AT-BSN (D)Learning Paradigm=Self-Supervised, Evaluation Protocol=Fully self-supervised, Model Variant=(D)2026.04 | 38.68 | 0.942 | |
| RA-Haar-tSVDMethod Category=Traditional + DNN2025.12 | 38.64 | 0.945 | |
| C-BSNLearning Paradigm=Self-Supervised, Evaluation Protocol=Fully self-supervised2026.04 | 38.6 | 0.941 | |
| SASLLearning Paradigm=Self-Supervised, Evaluation Protocol=Fully self-supervised2026.04 | 38.58 | 0.936 | |
| APR-RDMethod Category=DNN models (Self-supervised)2025.12 | 38.57 | 0.942 | |
| APR (RD)Learning Paradigm=Self-Supervised, Evaluation Protocol=Standard, Model Variant=(RD)2026.04 | 38.57 | 0.942 | |
| SDAPSources=ICCV232025.11 | 38.56 | 0.94 | |
| SDAPLearning Paradigm=Self-Supervised, Evaluation Protocol=Fully self-supervised2026.04 | 38.56 | 0.94 | |
| TM-BSN (s=4)Learning Paradigm=Self-Supervised, Evaluation Protocol=Fully self-supervised, Model Variant=s=42026.04 | 38.54 | 0.937 | |
| C-BSNLearning Paradigm=Self-Supervised, Evaluation Protocol=Standard2026.04 | 38.45 | 0.939 | |
| LG-BPNLearning Paradigm=Self-Supervised, Evaluation Protocol=Fully self-supervised2026.04 | 38.43 | 0.942 | |
| Ours(Full Pipeline)Training=Synthetic noise only, PD Strategy=true2019.04 | 38.4 | 0.945 | |
| Zhou et al.Training Supervision=Supervised (Synthetic pairs)2023.03 | 38.4 | 0.945 | |
| PDDType=Supervised, Train Data=Paired Noisy-Clean, #Param=0.71M2025.12 | 38.4 | 0.9434 | |
| NAFNetMethod Category=DNN models (Supervised)2025.12 | 38.36 | 0.943 | |
| AT-BSN (D)Learning Paradigm=Self-Supervised, Evaluation Protocol=Standard, Model Variant=(D)2026.04 | 38.29 | 0.939 | |
| A-Haar-tSVDMethod Category=Traditional + DNN2025.12 | 38.25 | 0.944 | |
| BNN (LAN)Training Supervision=Self-Supervised2023.03 | 38.18 | 0.938 | |
| SASLLearning Paradigm=Self-Supervised, Evaluation Protocol=Standard2026.04 | 38.18 | 0.938 | |
| Haar-tSVDMethod Category=Traditional denoisers2025.12 | 38.11 | 0.939 | |
| TM-BSN (s=4)Learning Paradigm=Self-Supervised, Evaluation Protocol=Standard, Model Variant=s=42026.04 | 38.1 | 0.935 | |
| AP-BSN+R3Training Supervision=Self-Supervised2023.03 | 38.09 | 0.937 | |
| AP-BSN + R3Learning Paradigm=Self-Supervised, Evaluation Protocol=Fully self-supervised2026.04 | 38.09 | 0.937 | |
| CBDNetAdditional training data=true2022.03 | 38.06 | 0.942 | |
| CBDNet2019.08 | 38.06 | 0.9421 | |
| CBDNetTraining Supervision=Supervised (Synthetic pairs)2023.03 | 38.05 | 0.942 | |
| CBDNetLearning Paradigm=Supervised2026.04 | 38.05 | 0.942 | |
| MSt-SVDMethod Category=Traditional denoisers2025.12 | 38.01 | 0.938 | |
| SASLMethod Category=DNN models (Self-supervised)2025.12 | 38.01 | 0.936 | |
| CBDNetType=Supervised, Train Data=Paired Noisy-Clean, #Param=6.79M2025.12 | 38 | 0.94 | |
| TWSCTraining=Synthetic noise only2019.04 | 37.93 | 0.94 | |
| Wu et al.Training Supervision=Unpaired2023.03 | 37.93 | 0.937 | |
| D-BSNType=Pseudo-Supervised, Train Data=Unpaired Noisy-Clean, #Param=6.62M2025.12 | 37.93 | 0.9373 | |
| DnCNN-B2019.08 | 37.9 | 0.943 | |
| TBSNType=Self-Supervised, Train Data=Only Noisy, #Param=12.74M2025.12 | 37.9 | 0.9288 | |
| DnCNNLearning Paradigm=Supervised2026.04 | 37.9 | 0.943 | |
| NSP(TBSN)Type=Self-Supervised, Train Data=Only Noisy, #Param=12.77M2025.12 | 37.87 | 0.9342 | |
| SDAPLearning Paradigm=Self-Supervised, Evaluation Protocol=Standard2026.04 | 37.86 | 0.937 | |
| BitonicMethod Category=Traditional denoisers2025.12 | 37.85 | 0.936 | |
| NSP(DBSN)Type=Self-Supervised, Train Data=Only Noisy, #Param=3.75M2025.12 | 37.8 | 0.9319 | |
| TBSNMethod Category=DNN models (Self-supervised)2025.12 | 37.79 | 0.94 | |
| CBM3DMethod Category=Traditional denoisers2025.12 | 37.73 | 0.934 | |
| SDAPType=Self-Supervised, Train Data=Only Noisy, #Param=3.66M2025.12 | 37.71 | 0.9278 | |
| FFDNet2019.08 | 37.61 | 0.9415 | |
| FFDNetMethod Category=DNN models (Supervised)2025.12 | 37.61 | 0.942 | |
| R2RType=Pseudo-Supervised, Train Data=Paired Noisy-Noisy, #Param=0.56M2025.12 | 37.61 | 0.9368 | |
| CBDNet(Syn)Training=Synthetic noise only2019.04 | 37.57 | 0.936 | |
| FFDNetTraining=Synthetic noise only, PD Strategy=true2019.04 | 37.56 | 0.931 | |
| AP-BSNType=Self-Supervised, Train Data=Only Noisy, #Param=3.66M2025.12 | 37.46 | 0.9244 | |
| MCWNNMTraining=Synthetic noise only2019.04 | 37.38 | 0.929 | |
| MCWNNMMethod Category=Traditional denoisers2025.12 | 37.38 | 0.929 | |
| C2NTraining Supervision=Unpaired2023.03 | 37.28 | 0.924 | |
| C2NType=Pseudo-Supervised, Train Data=Unpaired Noisy-Clean, #Param=217.26M2025.12 | 37.28 | 0.9237 | |
| C2NLearning Paradigm=Unpaired2026.04 | 37.28 | 0.924 | |
| TTT-MIMSources=ECCV242025.11 | 37.04 | — | |
| DnLUTSources=CVPR252025.11 | 36.67 | 0.922 | |
| CVF-SIDTraining Supervision=Self-Supervised2023.03 | 36.5 | 0.924 |