Research map / Image generation / Image restoration
RawMetaDiff: Unlocking Extreme Darkness from Dual-Exposure RAW with Meta-Guided Diffusion
Panjun Liu, Jiyuan Xia, Yuanshen Guan, Yong Li, Zhiqiang Lang, Ruikang Xu, Chang Chen, Dehua Song, Fenglong Song, Zhiwei Xiong
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 long-exposure conditioning (ablation) | RawMetaDiff · Synthetic ablation | PSNR: 21.37 -> 23.59 | Improved |
| add Meta-Normed Cross Attention (MNCA) (ablation) | RawMetaDiff · Synthetic ablation | PSNR: 21.42 -> 23.59 | Improved |
| add Meta-Assistant Color Transfer (MACT) (ablation) | RawMetaDiff · Synthetic ablation | Delta E: 7.7757 -> 5.5077 | Improved |
| add MNCA (ablation) | RawMetaDiff · Synthetic ablation | CLIP-IQA: 0.6533 -> 0.7008 | Improved |
| add MACT (ablation) | RawMetaDiff · Synthetic ablation | DeQA: 4.111 -> 4.115 | No clear effect |
| add long-exposure conditioning (ablation) | RawMetaDiff · Synthetic ablation | Delta E: 7.8052 -> 5.5077 | Improved |
| RawMetaDiff versus Restormer on synthetic data | RawMetaDiff · Synthetic low-light Raw | PSNR: 23.74 -> 23.74 | No clear effect |
| RawMetaDiff versus LSD2 on synthetic data | RawMetaDiff · Synthetic low-light Raw | MS-SSIM: 0.9031 -> 0.8745 | Hurt |
| RawMetaDiff versus HYPIR on real-world data | RawMetaDiff · DERaw real-world | DeQA: 2.017 -> 2.343 | Improved |
| naive two-frame concatenation for methods not designed for dual-frame input (Restormer) | Restormer · Synthetic low-light Raw | quality | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Restormer — Restormer: Efficient Transformer for High-Resolution Image Restoration (CVPR 2022) · better on 1 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
- Learning to See in the Extremely Dark (ICCV 2025 · similarity 0.89)
- From Abyssal Darkness to Blinding Glare: A Benchmark on Extreme Exposure Correction in Real World (ICCV 2025 · similarity 0.88)
- FreeAdapt: Unleashing Diffusion Priors for Ultra-High-Definition Image Restoration (ICLR 2026 · similarity 0.87)
- AGLLDiff: Guiding Diffusion Models Towards Unsupervised Training-free Real-world Low-light Image Enhancement (AAAI 2025 · similarity 0.87)
- Fourier Priors-Guided Diffusion for Zero-Shot Joint Low-Light Enhancement and Deblurring (CVPR 2024 · similarity 0.86)
- Diff-Shadow: Global-guided Diffusion Model for Shadow Removal (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.
