Research map / Image generation / Image restoration
Learning Physics-Informed Noise Models from Dark Frames for Low-Light Raw Image Denoising
Hansen Feng, Lizhi Wang, Yiqi Huang, Yuzhi Wang, Lin Zhu, Hua Huang
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 physics-guided noise decoupling (ablation) | PNNP · ELD Sony | PSNR: 44.26 -> 46.39 | Improved |
| add physics-guided noise decoupling (ablation) | PNNP · SID Sony | PSNR: 38.45 -> 40.83 | Improved |
| use 1x1 instead of 3x3 convolutions in PPM (ablation) | PNNP · ELD Sony | PSNR: 44.69 -> 46.39 | Improved |
| use 1x1 instead of 3x3 convolutions in PPM (ablation) | PNNP · SID Sony | PSNR: 39.68 -> 40.83 | Improved |
| use ISO-aware dual-branch structure instead of single branch (ablation) | PNNP · ELD Sony | PSNR: 45.89 -> 46.39 | Improved |
| use physics-aware PPM instead of flow-based modules (ablation) | PNNP · ELD Sony | PSNR: 46.00 -> 46.39 | Improved |
| add quantile loss to DDL (ablation) | PNNP · ELD Sony | PSNR: 46.05 -> 46.39 | Improved |
| add CDF loss to DDL (ablation) | PNNP · ELD Sony | PSNR: 46.16 -> 46.39 | Improved |
| add quantile loss to DDL (ablation) | PNNP · Pixel-wise noise | KLD: 0.00203 -> 0.00048 | Improved |
| learn noise model from dark frames (PNNP) instead of paired real data (Paired Data) | denoiser · ELD Sony | PSNR: 43.22 -> 46.39 | Improved |
| PNNP versus paired real data on high-quality LRID | denoiser · LRID-Indoor | PSNR: 44.60 -> 44.69 | No clear effect |
| PNNP versus paired real data on LRID-Outdoor | denoiser · LRID-Outdoor | PSNR: 44.33 -> 44.17 | No clear effect |
Closest papers by meaning
- A Physics-Based Noise Formation Model for Extreme Low-Light Raw Denoising (CVPR 2020 · similarity 0.90)
- Calibration-Free Raw Image Denoising via Fine-Grained Noise Estimation (TPAMI 2025 · similarity 0.90)
- Dark Noise Diffusion: Noise Synthesis for Low-Light Image Denoising (TPAMI 2025 · similarity 0.90)
- Rethinking Noise Synthesis and Modeling in Raw Denoising (ICCV 2021 · similarity 0.90)
- Noise Modeling in One Hour: Minimizing Preparation Efforts for Self-supervised Low-Light RAW Image Denoising (CVPR 2025 · similarity 0.88)
- Ditch the Denoiser: Emergence of Noise Robustness in Self-Supervised Learning from Data Curriculum (NeurIPS 2025 · similarity 0.88)
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.
