Research map / Image generation / Image restoration
UniRes: Universal Image Restoration for Complex Degradations
Mo Zhou, Keren Ye, Mauricio Delbracio, Peyman Milanfar, Vishal M. Patel, Hossein Talebi
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 multi-task training and weighted combination of four task experts (UniRes vs StableSR) | UniRes · DiversePhotos x1 | MUSIQ: 61.39 -> 68.22 | Improved |
| add multi-task training beyond super-resolution only (ablation) | UniRes · DiversePhotos x1 | MUSIQ: 47.76 -> 68.22 | Improved |
| combine tasks instead of single-task inference (denoise only) (ablation) | UniRes · DiversePhotos x1 | MUSIQ: 39.21 -> 68.22 | Improved |
| add DownLQ x4 term to the combination (ablation) | UniRes · DiversePhotos x1 | MUSIQ: 62.81 -> 68.19 | Improved |
| add blind restoration (BR) expert to the combination (ablation) | UniRes · DiversePhotos x1 | MUSIQ: 68.19 -> 68.22 | No clear effect |
| add super-resolution expert (ablation) | UniRes · DiversePhotos x1 | MUSIQ: 59.70 -> 62.81 | Improved |
| add denoise expert (ablation) | UniRes · DiversePhotos x1 | MUSIQ: 61.41 -> 62.81 | Improved |
| use DownLQ factor x4 instead of x2 (ablation) | UniRes · DiversePhotos x1 | MUSIQ: 55.38 -> 68.22 | Improved |
| widen search range from [0,1] to [-0.2,1.2] (ablation) | UniRes · DiversePhotos x1 | MUSIQ: 63.24 -> 68.22 | Improved |
| replace grid search with 8 most frequent weight sets (ablation) | UniRes · DiversePhotos x1 | MUSIQ: 68.22 -> 68.02 | No clear effect |
| replace grid search with Random Forest weight prediction (ablation) | UniRes · DiversePhotos x1 | MUSIQ: 68.22 -> 61.91 | Hurt |
| use UniRes instead of StableSR on RealSR x4 (single degradation) | UniRes · RealSR x4 | LPIPS: 0.3002 -> 0.3282 | Hurt |
Closest papers by meaning
- GenDeg: Diffusion-based Degradation Synthesis for Generalizable All-In-One Image Restoration (CVPR 2025 · similarity 0.90)
- Unsupervised Diffusion-Based Degradation Modeling for Real-World Super-Resolution (AAAI 2025 · similarity 0.89)
- DGSolver: Diffusion Generalist Solver with Universal Posterior Sampling for Image Restoration (NeurIPS 2025 · similarity 0.88)
- UniDemoiré: Towards Universal Image Demoiréing with Data Generation and Synthesis (AAAI 2025 · similarity 0.88)
- UP-Restorer: When Unrolling Meets Prompts for Unified Image Restoration (AAAI 2025 · similarity 0.88)
- Learning Where and What to Restore for Composite Image Restoration (NeurIPS 2026 · 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.
