Research map / Image generation / Image restoration
The Devil Is in the Dark Pixels: Toward Brightness Bias-Robust Denoising
Sungjun Cho, Zhuangzhuang Chen, Xiaomeng 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 |
|---|---|---|---|
| replace MSE with BBRD | NAFNet · SIDD | PSNR: 39.47 -> 40.12 (+0.65 dB) | Improved |
| replace MSE with BBRD | NAFNet · PolyU | PSNR: 38.50 -> 39.06 (+0.56 dB) | Improved |
| replace MSE with BBRD across 8 architectures | 8 backbones · SIDD and PolyU | PSNR-D, PSNR-B, PSNR, SSIM: SwinIR SIDD PSNR-D 37.29 -> 37.83 | Improved |
| use Heteroscedastic NLL instead of MSE | NAFNet · SIDD | PSNR-D: 37.51 -> 37.29 | Hurt |
| add per-band noise normalization before DRO (ablation) | BBRD · SIDD | PSNR-D: 36.67 -> 37.96 | Improved |
| add DRO on top of normalization (ablation) | BBRD · SIDD | PSNR-D: 36.82 -> 37.96 | Improved |
| use GMM bands instead of uniform quantile bands (ablation) | BBRD · SIDD | PSNR-D: 37.91 -> 37.96 | Improved |
| assign bands from clean target instead of noisy input (ablation) | BBRD · SIDD | PSNR-D: 37.43 -> 37.96 | Improved |
| use group-level DRO instead of per-pixel inverse variance weighting (ablation) | BBRD · SIDD | PSNR-D: 37.35 -> 37.96 | Improved |
| set DRO temperature eta=5 instead of 1 (ablation) | BBRD · SIDD | PSNR-D: 36.80 -> 37.96 | Improved |
| use K=5 bands instead of K=4 or K=8 (ablation) | BBRD · SIDD | PSNR-D: -0.24 dB (K=4), -0.30 dB (K=8) | Improved |
| add BBRD under signal-dependent noise | NAFNet · LumaSet-700 | PSNR-D: 26.50 -> 26.80 | Improved |
Closest papers by meaning
- Dark Noise Diffusion: Noise Synthesis for Low-Light Image Denoising (TPAMI 2025 · similarity 0.89)
- 2-Shots in the Dark: Low-Light Denoising with Minimal Data Acquisition (CVPR 2026 · similarity 0.89)
- Calibration-Free Raw Image Denoising via Fine-Grained Noise Estimation (TPAMI 2025 · similarity 0.88)
- Interpretable Unsupervised Joint Denoising and Enhancement for Real-World low-light Scenarios (ICLR 2025 · similarity 0.87)
- ZERO-IG: Zero-Shot Illumination-Guided Joint Denoising and Adaptive Enhancement for Low-Light Images (CVPR 2024 · similarity 0.87)
- Rethinking Reconstruction and Denoising in the Dark: New Perspective, General Architecture and Beyond (CVPR 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.
