Research map / Image generation / Image restoration
MoE-DiffIR: Task-customized Diffusion Priors for Universal Compressed Image Restoration
Yulin Ren, Xin Li, Bingchen Li, Xingrui Wang, Mengxi China Guo, Shijie Zhao, Li Zhang, Zhibo Chen
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 MoE-Prompt instead of soft-weighted multiple prompts (ablation) | MoE-DiffIR · LIVE1 seen tasks | LPIPS: 0.1212 -> 0.1179 | Improved |
| use MoE-Prompt instead of a single prompt (ablation) | MoE-DiffIR · LIVE1 seen tasks | LPIPS: 0.1272 -> 0.1179 | Improved |
| add prompts at all (MoE-Prompt vs no prompt) (ablation) | MoE-DiffIR · ICB unseen Cross Types | FID: 210.84 -> 176.87 | Improved |
| add Visual2Text adapter to MoE-Prompt (ablation) | MoE-DiffIR · LIVE1 (21 tasks) | LPIPS: 0.1179 -> 0.1145 | Improved |
| add degradation prior (DP) to router (ablation) | MoE-DiffIR · LIVE1 (21 tasks) | PSNR: 29.02 -> 29.07 | Improved |
| fine-tune VAE decoder with LQ information (decoder compensator) (ablation) | MoE-DiffIR · LIVE1 (avg 21 tasks) | PSNR: 22.70 -> 29.03 | Improved |
| train VAE from scratch instead of fine-tuning decoder (ablation) | MoE-DiffIR · LIVE1 (avg 21 tasks) | PSNR: 29.03 -> 25.12 | Hurt |
| select K=1 prompt instead of K=3 (ablation) | MoE-Prompt router · LIVE1 | LPIPS / PSNR | Mixed |
| use too many prompts (N above 7) (ablation) | MoE-Prompt · LIVE1 | PSNR/LPIPS | No clear effect |
| MoE-DiffIR versus StableSR on distortion | MoE-DiffIR · seven codecs average | PSNR: +0.41dB | Improved |
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.
- PromptIR — PromptIR: Prompting for All-in-One Image Restoration (NeurIPS 2023) · better on 2 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
- LucidFlux: Caption-Free Universal Image Restoration via a Large-Scale Diffusion Transformer (ICLR 2026 · similarity 0.90)
- Adapting Text-to-Image Generation with Feature Difference Instruction for Generic Image Restoration (CVPR 2025 · similarity 0.88)
- UP-Restorer: When Unrolling Meets Prompts for Unified Image Restoration (AAAI 2025 · similarity 0.87)
- Learning Diffusion Texture Priors for Image Restoration (CVPR 2024 · similarity 0.87)
- ReFIR: Grounding Large Restoration Models with Retrieval Augmentation (NeurIPS 2024 · similarity 0.87)
- Selective Hourglass Mapping for Universal Image Restoration Based on Diffusion Model (CVPR 2024 · similarity 0.87)
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.
