Research map / Image generation / Image restoration
Towards Illumination-Aware Restoration of Metalens-Captured Images: A New Dataset and a Strong Baseline
Fen Fang, Xinan Liang, Muli Yang, Jinghong Zheng, Tobias Mass, Ying Sun, Xulei Yang, Xuewu Xu, Zhengguo Li
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 Spatial-Aware Attention Scheduler (M1) (ablation) | DRMI · IlluMeta Setup 1 (avg) | PSNR (dB), whole: 16.8 -> 17.3 | Improved |
| add Spatial-Aware Attention Scheduler (M1) (ablation) | DRMI · IlluMeta Setup 1 (avg) | PSNR (dB), peripheral: 14.4 -> 15.3 | Improved |
| add Retraining-Free Illumination Adapter (M2) (ablation) | DRMI · IlluMeta Setup 1 (avg) | PSNR (dB), whole: 16.8 -> 18.2 | Improved |
| add both M1 and M2 (ablation) | DRMI · IlluMeta Setup 1 (avg) | PSNR (dB), whole: 18.2 -> 18.8 | Improved |
| add M1 on top of M2 for peripheral area (ablation) | DRMI · IlluMeta Setup 1 (avg) | PSNR (dB), peripheral: 16.5 -> 16.6 | No clear effect |
| add Attention Scheduler | DRMI · DIV2K-Meta whole image | PSNR (dB): 22.09 -> 22.17 | No clear effect |
| add Attention Scheduler | DRMI · DIV2K-Meta peripheral | PSNR (dB): 16.56 -> 17.46 | Improved |
| add Attention Scheduler | SFNet · DIV2K-Meta whole image | PSNR (dB): 18.22 -> 18.47 | No clear effect |
| train the restoration model jointly on multiple ILs instead of one IL (ablation) | restoration model · IlluMeta | PSNR (dB): 24.20 -> 22.71 | Hurt |
| add RL illumination adapter instead of joint multi-IL training | DRMI · IlluMeta Setup 1 (avg) | PSNR (dB): 16.8 -> 18.8 | Improved |
| add RL illumination adapter instead of joint multi-IL training | NAFNet · IlluMeta Setup 1 (avg) | PSNR (dB): 19.7 -> 19.9 | No clear effect |
| add RL illumination adapter instead of joint multi-IL training | HINet · IlluMeta Setup 1 (avg) | PSNR (dB): 19.9 -> 21.0 | 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.
- HINet — HiNet: Deep Image Hiding by Invertible Network (ICCV 2021) · better on 2 of 2 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
- After the Party: Navigating the Mapping From Color to Ambient Lighting (ICCV 2025 · similarity 0.87)
- Robust Low-light Scene Restoration via Illumination Transition (ICCV 2025 · similarity 0.87)
- Learning to See in the Extremely Dark (ICCV 2025 · similarity 0.87)
- Luminance-Aware Statistical Quantization: Unsupervised Hierarchical Learning for Illumination Enhancement (NeurIPS 2025 · similarity 0.86)
- UARE: A Unified Vision-Language Model for Image Quality Assessment, Restoration, and Enhancement (CVPR 2026 · similarity 0.86)
- UniLumos: Fast and Unified Image and Video Relighting with Physics-Plausible Feedback (NeurIPS 2025 · similarity 0.85)
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.
