Research map / Image generation / Image restoration
UniRestorer: Universal Image Restoration via Adaptively Estimating Image Degradation at Proper Granularity
Jingbo Lin, Zhilu Zhang, Wenbo Li, Renjing Pei, Hang Xu, Hongzhi Zhang, Wangmeng Zuo
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 |
|---|---|---|---|
| use multi-granularity MoE with degradation and granularity estimation instead of prior all-in-one methods | UniRestorer · All-in-one 7 tasks average | PSNR: 28.27 -> 31.34 | Improved |
| use UniRestorer vs NAFNet on LIVE1 compression artifact removal | UniRestorer · LIVE1 CAR | PSNR: 31.26 -> 30.57 | Hurt |
| add granularity router G_g on top of degradation router G_d (ablation) | UniRestorer · Mixed-degradation Urban100 out-dist | PSNR: 18.60 -> 19.45 | Improved |
| add L_dg uncertainty loss and L_load load-balance loss (ablation) | UniRestorer routers · Mixed-degradation Urban100 out-dist | PSNR: 18.37 -> 19.45 | Improved |
| use multiple granularities {1,4,8} instead of single level {8} (ablation) | UniRestorer · Mixed-degradation Urban100 | PSNR out-dist: 18.35 -> 19.45 | Improved |
| add a fourth granularity level {1,4,8,16} (ablation) | UniRestorer · Mixed-degradation Urban100 | PSNR out-dist: 19.45 -> 19.47 | No clear effect |
| increase number of DR clusters from 1 to 16 (ablation) | UniRestorer single-level · Mixed-degradation Urban100 | PSNR in-dist: 22.06 -> 24.30 | Improved |
| increase number of DR clusters from 4 to 16 (ablation) | UniRestorer single-level · Mixed-degradation Urban100 out-dist | PSNR: 18.49 -> 18.44 | No clear effect |
| use fine-grained DR extractor instead of DA-CLIP for clustering (ablation) | UniRestorer · Rain200H | PSNR: 31.65 -> 32.01 | Improved |
| cluster data with learned extractor rather than manual dataset-source split (ablation) | UniRestorer · Rain200L | PSNR: 41.43 -> 41.61 | Improved |
| use instruction mode against single-task models | UniRestorer · Single-task comparison, LIVE1 q=10 | PSNR: 27.95 -> 27.94 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Restormer — Restormer: Efficient Transformer for High-Resolution Image Restoration (CVPR 2022) · better on 3 of 3 shared results
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
- UniRes: Universal Image Restoration for Complex Degradations (ICCV 2025 · similarity 0.92)
- Learning Where and What to Restore for Composite Image Restoration (NeurIPS 2026 · similarity 0.87)
- Continuous Expert Assembly: Instance-Conditioned Low-Rank Residuals for All-in-One Image Restoration (NeurIPS 2026 · similarity 0.87)
- UniDemoiré: Towards Universal Image Demoiréing with Data Generation and Synthesis (AAAI 2025 · similarity 0.87)
- Test-Time Degradation Adaptation for Open-Set Image Restoration (ICML 2024 · similarity 0.86)
- UP-Restorer: When Unrolling Meets Prompts for Unified Image Restoration (AAAI 2025 · similarity 0.86)
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.
