Research map / Image generation / Image restoration
Learning Camera-Aware Noise Models
Ke-Chi Chang, Ren Wang, Hung-Jin Lin, Yu-Lun Liu, Chia-Ping Chen, Yu-Lin Chang, Hwann-Tzong Chen
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 feature matching loss (ablation) | CA-NoiseGAN · SIDD noise modeling | KL divergence: 0.01445 -> 0.01374 | Improved |
| add Camera-Encoding Network without triplet loss (ablation) | CA-NoiseGAN · SIDD noise modeling | KL divergence: 0.01374 -> 0.01412 | No clear effect |
| add triplet loss to the Camera-Encoding Network (ablation) | CA-NoiseGAN · SIDD noise modeling | KL divergence: 0.01412 -> 0.00159 | Improved |
| use Poisson-Gaussian instead of Gaussian as initial noise (ablation) | CA-NoiseGAN · SIDD noise modeling | KL divergence: 0.06265 -> 0.00159 | Improved |
| use CA-NoiseGAN instead of Noise Flow | DnCNN denoiser · SIDD real-world denoising (500k) | PSNR: 48.52 -> 48.71 | Improved |
| use CA-NoiseGAN instead of Noise Flow | noise model · SIDD noise modeling | KL divergence: 0.00912 -> 0.00159 | Improved |
| add real data to learned-noise-model synthetic training | DnCNN denoiser · SIDD real-world denoising (500k) | PSNR: 48.71 -> 48.72 | No clear effect |
| increase noise model training data from 100k to 500k | CA-NoiseGAN · SIDD real-world denoising | PSNR: 47.97 -> 48.71 | Improved |
| add Camera-Encoding Network with triplet loss (ablation) | camera-specific denoisers · SIDD camera IP | PSNR: 57.4672 -> 58.6073 | Improved |
| add Camera-Encoding Network with triplet loss (ablation) | camera-specific denoisers · SIDD camera N6 | PSNR: 44.7954 -> 46.4726 | Improved |
| add Camera-Encoding Network without triplet loss (ablation) | camera-specific denoisers · SIDD cameras (IP, GP, S6, N6, G4) | PSNR: 57.4672 -> 49.8788 (IP); 44.7954 -> 41.8447 (N6) | Mixed |
| use latent vector from a different camera | CA-NoiseGAN · SIDD noise modeling | KL divergence: 0.00159 -> 0.01324 | Hurt |
Closest papers by meaning
- Noise2NoiseFlow: Realistic Camera Noise Modeling Without Clean Images (CVPR 2022 · similarity 0.90)
- Estimating Fine-Grained Noise Model via Contrastive Learning (CVPR 2022 · similarity 0.89)
- Rethinking Noise Synthesis and Modeling in Raw Denoising (ICCV 2021 · similarity 0.89)
- Calibration-Free Raw Image Denoising via Fine-Grained Noise Estimation (TPAMI 2025 · similarity 0.88)
- sRGB Real Noise Modeling via Noise-Aware Sampling with Normalizing Flows (ICLR 2024 · similarity 0.88)
- Realistic Noise Synthesis with Diffusion Models (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.
