Research map / Label-efficient and robust learning / Self-supervised and contrastive
Noise2Same: Optimizing A Self-Supervised Bound for Image Denoising
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 |
|---|---|---|---|
| replace blind-spot masked loss with Noise2Same reconstruction plus invariance loss | self-supervised denoising · ImageNet | PSNR: 21.36 -> 22.26 | Helped |
| replace blind-spot masked loss with Noise2Same | self-supervised denoising · HanZi | PSNR: 13.94 -> 14.38 | Helped |
| replace blind-spot masked loss with Noise2Same | self-supervised denoising · Planaria C1 | PSNR: 27.58 -> 29.48 | Helped |
| use Noise2Same instead of Noise2Self-Donut | self-supervised denoising · BSD68 | PSNR: 28.20 -> 27.95 | Hurt |
| use strictly J-invariant shifting RF instead of mask-based Gaussian replacement | blind-spot denoising · BSD68 | PSNR: 26.14 -> 27.15 | Helped |
| use UPS replacement (less strict J-invariance) instead of strictly J-invariant shifting RF | blind-spot denoising · BSD68 | PSNR: 27.15 -> 27.71 | Helped |
| set sigma_loss in invariance weight close to true noise sigma (ablation) | Noise2Same · HanZi with additive Gaussian noise | PSNR | Helped |
| self-supervised vs supervised Noise2True | Noise2Same · BSD68 | PSNR: 29.06 -> 27.95 | Hurt |
Closest papers by meaning
- Positive2Negative: Breaking the Information-Lossy Barrier in Self-Supervised Single Image Denoising (CVPR 2025 · similarity 0.91)
- Noise2Score: Tweedie’s Approach to Self-Supervised Image Denoising without Clean Images (NeurIPS 2021 · similarity 0.91)
- Zero-Shot Noise2Mean: Gap Minimization for Efficient Denoising from a Single Noisy Image (AAAI 2025 · similarity 0.88)
- Ditch the Denoiser: Emergence of Noise Robustness in Self-Supervised Learning from Data Curriculum (NeurIPS 2025 · similarity 0.88)
- UNSURE: self-supervised learning with Unknown Noise level and Stein's Unbiased Risk Estimate (ICLR 2025 · similarity 0.87)
- Exploring Efficient Asymmetric Blind-Spots for Self-Supervised Denoising in Real-World Scenarios (CVPR 2024 · similarity 0.87)
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.