Research map / Image generation / Image restoration
A High-Quality Denoising Dataset for Smartphone Cameras
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train DnCNN on SIDD ground truth with real noisy images instead of post-processed low-ISO ground truth with real noisy images (ablation) | DnCNN · SIDD subset A and subset B | PSNR | Helped |
| Train DnCNN with real noise rather than synthetic Gaussian noise (ablation) | DnCNN trained on SIDD ground truth · SIDD subset A and subset B | PSNR | Helped |
| Train DnCNN with SIDD ground truth instead of post-processed low-ISO ground truth, using synthetic noise (ablation) | DnCNN · SIDD subset A and subset B | PSNR | No clear effect |
| Use dense local alignment instead of global 2D translation alignment (ablation) | SIDD ground-truth estimation pipeline · Synthetic image sequences with smartphone-like local shifts | PSNR | Helped |
| Estimate clipped-pixel means with WLS fitting of the CDF rather than MLE with censoring (ablation) | SIDD ground-truth estimation pipeline · Synthetic images with noise and intensity clipping | MSE | Helped |
| Train DnCNN on SIDD ground truth with real noise rather than on low-ISO ground truth with real noise (ablation) | DnCNN · SIDD subset A | noise estimate σ: 8.02 × 10−1 -> 4.62 × 10−1 | Helped |
| Train DnCNN on SIDD ground truth with real noise rather than on low-ISO ground truth with real noise (ablation) | DnCNN · SIDD subset B | noise estimate σ: 1.04 -> 4.18 × 10−1 | Helped |
| Denoise in raw-RGB rather than sRGB | Benchmarked denoising algorithms · SIDD 40-image benchmark | PSNR | Mixed |
| Train DnCNN on real SIDD images and ground truth rather than use the existing trained model (ablation) | DnCNN · SIDD subset A and subset B | PSNR | Helped |
| Use the proposed SIDD ground truth rather than DND post-processing for ground-truth estimation | Ground-truth image estimation · Five compared image pairs | noise estimate σ | Helped |
Closest papers by meaning
- Real-World Mobile Image Denoising Dataset with Efficient Baselines (CVPR 2024 · similarity 0.92)
- NERDS: A General Framework to Train Camera Denoisers from Raw-RGB Noisy Image Pairs (ICLR 2023 · similarity 0.90)
- IDR: Self-Supervised Image Denoising via Iterative Data Refinement (CVPR 2022 · similarity 0.88)
- Joint HDR Denoising and Fusion: A Real-World Mobile HDR Image Dataset (CVPR 2023 · similarity 0.88)
- Benchmarking Denoising Algorithms With Real Photographs (CVPR 2017 · similarity 0.88)
- SIDL: A Real-World Dataset for Restoring Smartphone Images with Dirty Lenses (AAAI 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.