Research map / Image generation / Image restoration
Multi-view Self-supervised Disentanglement for General Image Denoising
Hao Chen, Chenyuan Qu, Yu Zhang, Chen Chen, Jianbo Jiao
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 Bernoulli manifold mixture loss (weight 0.025 vs 0.000) (ablation) | MeD · CBSD68 Gaussian sigma=75 | PSNR: 23.78 -> 26.01 | Improved |
| raise manifold loss weight from 0.025 to 0.050 (ablation) | MeD · CBSD68 Gaussian sigma=75 | PSNR: 26.01 -> 24.29 | Hurt |
| increase views from 2 to 4 (ablation) | MeD · Noise Pool Speckle | PSNR: 37.21 -> 37.47 | Improved |
| increase views from 2 to 4 (ablation) | MeD · training cost | training time: 10x training time | Hurt |
| pretrain with MeD instead of N2C (ablation) | N2C fine-tuning · CBSD68 unseen noise average | PSNR: 33.50 -> 34.46 | Improved |
| pretrain with MeD instead of N2N (ablation) | N2N fine-tuning · CBSD68 unseen noise average | PSNR: 33.59 -> 34.00 | Improved |
| pretrain with MeD instead of LIR (ablation) | LIR fine-tuning · CBSD68 unseen noise average | PSNR: 28.38 -> 29.73 | Improved |
| train with random sigma instead of fixed sigma=25 (ablation) | MeD · CBSD68 Gaussian sigma=75 | PSNR: 23.09 -> 25.40 | Improved |
| train MeD on Noise Pool rather than using supervised N2C | MeD vs N2C · CBSD68 Noise Pool average | PSNR: 34.08 -> 36.02 | Improved |
| use MeD on low noise level with fixed sigma 25 training | MeD vs DBD4 · CBSD68 sigma=15 | PSNR: 33.57 -> 33.11 | Hurt |
Closest papers by meaning
- Iterative Denoiser and Noise Estimator for Self-Supervised Image Denoising (ICCV 2023 · similarity 0.91)
- High-Quality Self-Supervised Deep Image Denoising (NeurIPS 2019 · similarity 0.90)
- Unpaired Learning of Deep Image Denoising (ECCV 2020 · similarity 0.90)
- Positive2Negative: Breaking the Information-Lossy Barrier in Self-Supervised Single Image Denoising (CVPR 2025 · similarity 0.88)
- Zero-Shot Noise2Mean: Gap Minimization for Efficient Denoising from a Single Noisy Image (AAAI 2025 · similarity 0.88)
- APR-RD: Complemental Two Steps for Self-Supervised Real Image Denoising (AAAI 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.
