Image Restoration on All-in-one multi-degradation dataset (H, R, N combinations) benchmark leaderboard · SOTA2 Research

Image Restoration on All-in-one multi-degradation dataset (H, R, N combinations)

30.45PSNR (H+R+N)MLMIR

Evaluation Results

Method	Links
MLMIR2026.06		30.45	90.1	32.52	96.5	28.73	87.6	32.23	91.7	31.82	97.9	40.11	98.7	32.72	92.9	32.65	93.6
Perceive-IR2026.06		30.22	89.4	31.55	94.5	28.32	87.1	30.52	84.9	31.9	98	38.95	98.3	31.88	92.3	31.9	92.1
DSwinIR2026.06		30.09	89.9	31.07	96	28.45	87.3	31.29	91.5	31.77	97.3	39.09	98.2	32.22	92.3	31.99	93.2
FDTANet2026.06		29.68	84.6	30.91	93.7	29.06	87.4	27.63	82.7	31.91	98.1	29.13	89.2	28.88	85.7	29.6	88.8
Defusion2026.06		29.66	84.9	29.93	93.7	28.11	86.6	32.17	92	31.29	96.9	40.15	98.6	32.72	92.5	32.01	92.2
VLU-Net2026.06		29.35	84.2	30.38	93.5	28.79	86.9	28.21	84.3	31.37	95.9	33.82	96.8	27.88	85.3	29.97	89.6
Ref-IRT2026.06		28.95	82.3	32.33	96.1	28.62	87.2	31.22	83.1	30.69	96.1	31.89	90.2	29.39	89.9	30.44	89.3