Research map / Learning theory and algorithms / Optimal transport
Degradation-Aware Residual-Conditioned Optimal Transport for Unified Image Restoration
Xiaole Tang, Xiang Gu, Xiaoyi He, Xin Hu, Jian Sun
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 |
|---|---|---|---|
| condition second pass on multi-scale residual embeddings instead of nothing (ablation) | DA-RCOT · Five-degradation All-in-One | average PSNR: 28.39 -> 30.40 | Improved |
| condition on intermediate restored result x0 instead of nothing (ablation) | DA-RCOT · Five-degradation All-in-One | average PSNR: 28.39 -> 28.45 | No clear effect |
| condition on multi-scale residuals instead of single R0 (ablation) | DA-RCOT · Five-degradation All-in-One | average PSNR: 29.81 -> 30.40 | Improved |
| add residual regularization g(r) to transport cost (ablation) | DA-RCOT · Five-degradation All-in-One | average PSNR: 30.05 -> 30.40 | Improved |
| add task loss to unpaired objective (ablation) | DA-RCOT unpaired · Three-degradation average | PSNR: 29.28 -> 30.20 | Improved |
| add task loss to paired FROT objective (ablation) | DA-RCOT paired · Three-degradation average | PSNR: 32.35 -> 32.60 | Improved |
| use FROT objective instead of L1 distance (ablation) | DA-RCOT paired · Three-degradation average | PSNR: 31.82 -> 32.35 | Improved |
| add REC to MPRNet (ablation) | MPRNet · Rain100L | PSNR: 34.95 -> 36.78 | Improved |
| add REC to NAFNet (ablation) | NAFNet · Rain100L | PSNR: 35.58 -> 37.10 | Improved |
| add REC to Restormer (ablation) | Restormer · Rain100L | PSNR: 36.74 -> 38.22 | Improved |
| add degradation-aware residual conditioning to RCOT | RCOT · Three-degradation average (paired) | PSNR: 32.13 -> 32.60 | Improved |
| DA-RCOT vs PromptIR under unseen noise sigma=75 | PromptIR · OOD denoising BSD68 sigma=75 | PSNR: 19.23 -> 21.64 | Improved |
Closest papers by meaning
- Residual-Conditioned Optimal Transport: Towards Structure-Preserving Unpaired and Paired Image Restoration (ICML 2024 · similarity 0.95)
- Bend the Basics: Degradation-Aware Deformable Tokenization for All-in-One Image Restoration (ICML 2026 · similarity 0.90)
- Visual-Instructed Degradation Diffusion for All-in-One Image Restoration (CVPR 2025 · similarity 0.90)
- Learning Where and What to Restore for Composite Image Restoration (NeurIPS 2026 · similarity 0.89)
- Continuous Expert Assembly: Instance-Conditioned Low-Rank Residuals for All-in-One Image Restoration (NeurIPS 2026 · similarity 0.88)
- GenDeg: Diffusion-based Degradation Synthesis for Generalizable 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.
