Research map / Image generation / Image restoration
Noise2Score: Tweedie’s Approach to Self-Supervised Image Denoising without Clean Images
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 |
|---|---|---|---|
| Use AR-DAE score estimation followed by Gaussian Tweedie denoising instead of SURE | Self-supervised image denoising · BSD68, Gaussian σ=25, known parameters | PSNR (dB): 28.78 -> 29.12 | Helped |
| Use AR-DAE score estimation followed by Gaussian Tweedie denoising instead of Noise2Void | Self-supervised image denoising · BSD68, Gaussian σ=25, known parameters | PSNR (dB): 26.77 -> 29.12 | Helped |
| Use AR-DAE score estimation followed by Poisson Tweedie denoising instead of PURE | Self-supervised image denoising · BSD68, Poisson ζ=0.01, known parameters | PSNR (dB): 30.16 -> 30.81 | Helped |
| Use AR-DAE score estimation followed by Gamma Tweedie denoising instead of Noise2Self | Self-supervised image denoising · BSD68, Gamma k=100, known parameters | PSNR (dB): 30.49 -> 32.67 | Helped |
| Use Noise2Score instead of Noise2Same | Self-supervised image denoising · Confocal MICE, real mixed Poisson-Gaussian noise | PSNR (dB): 36.42 -> 36.73 | Helped |
| Use Noise2Score rather than Noise2Void when training on Poisson noise ζ=0.01 and testing at varying Poisson levels | Self-supervised image denoising · BSD68, inference ζ=0.01 to 0.05 | PSNR (dB): 28.73 vs 30.81; 26.67 vs 29.01; 24.79 vs 28.00; 23.39 vs 27.25; 22.32 vs 26.64 | Helped |
Closest papers by meaning
- IDR: Self-Supervised Image Denoising via Iterative Data Refinement (CVPR 2022 · similarity 0.92)
- Noise Distribution Adaptive Self-Supervised Image Denoising Using Tweedie Distribution and Score Matching (CVPR 2022 · similarity 0.92)
- Convexity-Aware Noise Calibration: A Self-Supervised Framework for Noise-Level-Unknown Image Denoising (CVPR 2026 · similarity 0.92)
- Noise2Same: Optimizing A Self-Supervised Bound for Image Denoising (NeurIPS 2020 · similarity 0.91)
- Zero-Shot Noise2Mean: Gap Minimization for Efficient Denoising from a Single Noisy Image (AAAI 2025 · similarity 0.89)
- Positive2Negative: Breaking the Information-Lossy Barrier in Self-Supervised Single Image Denoising (CVPR 2025 · similarity 0.89)
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.