Research map / Image generation / Image restoration
Rethinking Expressivity and Degradation-Awareness in Attention for All-in-One Blind Image Restoration
Bin Ren, Runyi Yang, Qi Ma, Xu Zheng, Mengyuan Liu, Danda Pani Paudel, Luc Van Gool, Rita Cucchiara, Nicu Sebe
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add nonlinear value transform (ablation) | prompt-free PromptIR · 3-degradation average | PSNR: 30.75 -> 32.54 | Improved |
| add global spatial token (GST) (ablation) | prompt-free PromptIR · 3-degradation average | PSNR: 30.75 -> 32.67 | Improved |
| combine nonlinear value and GST (ablation) | prompt-free PromptIR · 3-degradation average | PSNR: 30.75 -> 32.96 | Improved |
| use residual form of nonlinear value instead of in-place (GELU) (ablation) | ExDA-Tiny · 3-degradation average | PSNR: 32.45 -> 32.71 | Improved |
| use residual form instead of in-place (Sigmoid) (ablation) | ExDA-Tiny · 3-degradation average | PSNR: 32.30 -> 32.63 | Improved |
| use learnable parametric mapping instead of non-parametric GELU (ablation) | ExDA-Tiny · 3-degradation average | PSNR: 32.30 -> 32.71 | Improved |
| apply nonlinear values in both encoder and decoder instead of encoder-only (ablation) | ExDA-Tiny · 3-degradation average | PSNR: 32.65 -> 32.71 | Improved |
| apply nonlinear values in both encoder and decoder instead of decoder-only (ablation) | ExDA-Tiny · 3-degradation average | PSNR: 32.60 -> 32.71 | Improved |
| use GST stride 2 instead of 1 (ablation) | ExDA-Tiny · 3-degradation average | PSNR: 32.45 -> 32.71 | Improved |
| use GST stride 2 instead of 4 or 8 (ablation) | ExDA-Tiny · 3-degradation average | PSNR: 32.52 and 32.40 -> 32.71 | Improved |
| add GST to improve degradation separability (ablation) | ExDA · embedding clustering | NMI / ARI: NMI 0.71 -> 0.88; ARI 0.56 -> 0.89 | Improved |
| use nonlinear value instead of linear value in synthetic restoration (ablation) | attention · MNIST restoration | PSNR: 19.2 -> 25.1 dB | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- PromptIR — PromptIR: Prompting for All-in-One Image Restoration (NeurIPS 2023) · better on 4 of 4 shared results
- AdaIR — AdaIR: Adaptive All-in-One Image Restoration via Frequency Mining and Modulation (ICLR 2025) · better on 3 of 3 shared results
- OneRestore — OneRestore: A Universal Restoration Framework for Composite Degradation (ECCV 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Sharing Key Semantics in Transformer Makes Efficient Image Restoration (NeurIPS 2024 · similarity 0.89)
- Visual-Instructed Degradation Diffusion for All-in-One Image Restoration (CVPR 2025 · similarity 0.89)
- Debiased All-in-one Image Restoration with Task Uncertainty Regularization (AAAI 2025 · similarity 0.89)
- Continuous Expert Assembly: Instance-Conditioned Low-Rank Residuals for All-in-One Image Restoration (NeurIPS 2026 · similarity 0.88)
- UP-Restorer: When Unrolling Meets Prompts for Unified Image Restoration (AAAI 2025 · similarity 0.88)
- Degradation-Aware Feature Perturbation for All-in-One Image Restoration (CVPR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
