Research map / Image generation / Image restoration
Learning Efficient and Effective Trajectories for Differential Equation-Based Image Restoration
Zhiyu Zhu, Jinhui Hou, Hui Liu, Huanqiang Zeng, Junhui Hou
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 acceleration distillation (NFE=2) on top of RL-reinforced ODE (NFE=10) (ablation) | GSAD-based IR diffusion · UIEBD | PSNR: 25.08 -> 26.30 | Improved |
| distill to 1 step instead of 2 steps (ablation) | GSAD-based IR diffusion · UIEBD | PSNR: 26.30 -> 26.25 | No clear effect |
| distill to 1 step instead of 2 steps (ablation) | GSAD-based IR diffusion · C60 | MUSIQ: 47.019 -> 44.583 | Hurt |
| distill reinforced ODE to 1 step (ablation) | GSAD-based IR diffusion · LOL | PSNR: 28.581 -> 28.184 | Hurt |
| distill reinforced ODE to 2 steps (ablation) | GSAD-based IR diffusion · LOL-v2 | PSNR: 29.535 -> 29.956 | Improved |
| distill reinforced ODE to 1 step (ablation) | GSAD-based IR diffusion · LOL | runtime (s): 0.81 -> 0.06 | Improved |
| distill reinforced ODE to 2 steps (ablation) | GSAD-based IR diffusion · Rain100H | PSNR: 34.77 -> 35.80 | Improved |
| distill reinforced ODE to 2 steps (ablation) | GSAD-based IR diffusion · Rain200H | PSNR: 30.49 -> 30.33 | Hurt |
| use trajectory-augmented model instead of GSAD | GSAD · Rain100L | PSNR: 42.57 -> 44.38 | Improved |
| use trajectory-augmented model instead of GSAD | GSAD · Rain200H | PSNR: 28.60 -> 30.33 | Improved |
| use trajectory-augmented model instead of transformer FADformer | FADformer · Rain200H | PSNR: 32.48 -> 30.33 | Hurt |
| use trajectory-augmented model instead of HCLR-net | HCLR-net · UIEBD | PSNR: 25.00 -> 26.30 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- AutoDIR — AutoDIR: Automatic All-in-One Image Restoration with Latent Diffusion (ECCV 2024) · better on 0 of 1 shared result
- Retinexformer — Retinexformer: One-stage Retinex-based Transformer for Low-light Image Enhancement (ICCV 2023) · 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
- Improving Diffusion-Based Image Restoration with Error Contraction and Error Correction (AAAI 2024 · similarity 0.89)
- UP-Restorer: When Unrolling Meets Prompts for Unified Image Restoration (AAAI 2025 · similarity 0.88)
- Reconciling Stochastic and Deterministic Strategies for Zero-shot Image Restoration using Diffusion Model in Dual (CVPR 2025 · similarity 0.88)
- Blind Image Restoration via Fast Diffusion Inversion (NeurIPS 2024 · similarity 0.88)
- Learning Diffusion Texture Priors for Image Restoration (CVPR 2024 · similarity 0.88)
- Image Restoration by Denoising Diffusion Models with Iteratively Preconditioned Guidance (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.
