Defocus-to-Focus SEM Restoration on Real defocused SEM images S→R protocol (Img2)
11.7CDMRN
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MRNArchitecture=UNet2026.04 | 11.7 | 0.6 | 1.4 | 1.1 | 1.4 | |
| Noise2Void (N2V)Architecture=UNet2026.04 | 10.8 | 0.8 | 1.4 | 1.1 | 1.4 | |
| Richardson–LucyMethod Type=Classical Deconvolution2026.04 | 10.7 | 0.8 | 1.4 | 1.1 | 1.4 | |
| Noise2Noise (N2N)Architecture=UNet2026.04 | 10.6 | 0.8 | 1.4 | 1.1 | 1.4 | |
| Defocused InputInput Type=Noisy/Defocused2026.04 | 10.4 | 0.7 | 2.5 | 1.9 | 2.5 | |
| BM3DMethod Type=Denoising2026.04 | 10.2 | 0.7 | 2.3 | 1.7 | 2.3 | |
| ViT-MAE (SEM)Pre-training Dataset=SEM, Architecture=ViT-MAE2026.04 | 10.1 | 0.6 | 2 | 1.5 | 2 | |
| Focused InputInput Type=Reference2026.04 | 9.7 | 0.6 | 2.2 | 1.6 | 2.2 | |
| ViT-MAE + MoE (ℓ1)Architecture=ViT-MAE + MoE, Loss=ℓ12026.04 | 9.7 | 0.8 | 2 | 1.6 | 2 | |
| ViT-MAE + MoE (ℓ1+TV)Architecture=ViT-MAE + MoE, Loss=ℓ1 + TV2026.04 | 9.4 | 0.8 | 1.3 | 1 | 1.3 |