Research map / Image generation / Image restoration
Self-Supervised Image Prior Learning With GMM From a Single Noisy Image
Haosen Liu, Xuan Liu, Jiangbo Lu, Shan Tan
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 |
|---|---|---|---|
| learn the GMM prior from a single noisy image with SS-GMM instead of supervised EPLL training on clean images | GMM image prior · Set12 | PSNR: over 0.3dB on average | Improved |
| SS-GMM versus EPLL on BSD68, whose EPLL model was trained on same-dataset images | GMM image prior · BSD68 | PSNR: 31.22 -> 31.26 at sigma=15 | No clear effect |
| use SS-GMM instead of BM3D on BSD68 | denoising · BSD68 | PSNR: about 0.1dB-0.2dB | Improved |
| use SS-GMM instead of BM3D on Set12 | denoising · Set12 | PSNR: 32.12 -> 32.18 at sigma=15; 29.73 -> 29.68 at sigma=25 | Mixed |
| use SS-GMM instead of NL-Bayes | denoising · Set12 and BSD68 | PSNR: 31.98 -> 32.18 on Set12 sigma=15 | Improved |
| use SS-GMM instead of DIP | denoising · BSD68 | PSNR: 28.00 -> 28.73 at sigma=25 | Improved |
| use SS-GMM instead of Self2Self at low noise | denoising · BSD68 | PSNR: 30.26 -> 31.26 at sigma=15 | Improved |
| use SS-GMM instead of Self2Self at high noise | denoising · BSD68 | PSNR: 25.92 -> 25.70 at sigma=50 | Hurt |
| replace the noise estimator of Chen et al. with the peak-region estimator | noise level estimation · Set12 | Bias, Std, sqrt(MSE) | Improved |
Closest papers by meaning
- Analyzing the Sample Complexity of Self-Supervised Image Reconstruction Methods (NeurIPS 2023 · similarity 0.87)
- Robust Test-Time Adaptation for Single Image Denoising Using Deep Gaussian Prior (ICCV 2025 · similarity 0.87)
- UNSURE: self-supervised learning with Unknown Noise level and Stein's Unbiased Risk Estimate (ICLR 2025 · similarity 0.87)
- Ditch the Denoiser: Emergence of Noise Robustness in Self-Supervised Learning from Data Curriculum (NeurIPS 2025 · similarity 0.86)
- Sublinear Variational Optimization of Gaussian Mixture Models with Millions to Billions of Parameters (NeurIPS 2026 · similarity 0.86)
- Rotation-Equivariant Self-Supervised Method in Image Denoising (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.
