Research map / Image generation / Image restoration
Spatially Adaptive Self-Supervised Learning for Real-World 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 |
|---|---|---|---|
| Use spatially adaptive BNN and LAN supervision instead of AP-BSN+R^3 | real-world sRGB image denoising · SIDD Validation | PSNR: 36.74 -> 37.39 | Helped |
| Use spatially adaptive BNN and LAN supervision instead of AP-BSN+R^3 | real-world sRGB image denoising · SIDD Benchmark | PSNR: 36.91 -> 37.41 | Helped |
| Use spatially adaptive BNN and LAN supervision instead of AP-BSN+R^3 | real-world sRGB image denoising · DND Benchmark | PSNR: 38.09 -> 38.18 | Helped |
| Add LAN supervision to BNN supervision (ablation) | U-Net denoiser on SIDD Validation · SIDD Validation | PSNR: 36.84 -> 37.30 | Helped |
| Add BNN supervision to LAN supervision (ablation) | U-Net denoiser on SIDD Validation · SIDD Validation | PSNR: 35.95 -> 37.30 | Helped |
| Add adaptive coefficients to balance BNN and LAN supervision by local flatness (ablation) | U-Net denoiser on SIDD Validation · SIDD Validation | PSNR: 37.30 -> 37.39; 0.09dB | Helped |
| Use BNN and LAN outputs as training supervision rather than spatially adaptive image-level fusion (ablation) | real-world image denoising · SIDD Validation | PSNR: 36.84 -> 37.39; 0.55dB | Helped |
| Replace image-level fusion with the trained denoising network at inference (ablation) | real-world image denoising · SIDD Validation | Time: 22.9 -> 4.8 ms | Helped |
| Increase BNN blind-neighborhood size from 1×1 toward 9×9 (ablation) | BNN on SIDD Validation · SIDD Validation | PSNR | Helped |
| Increase BNN blind-neighborhood size beyond 9×9 (ablation) | BNN on SIDD Validation · SIDD Validation | PSNR | Hurt |
| Increase LAN local receptive size from 1×1 (ablation) | LAN and final denoiser on SIDD Validation · SIDD Validation | PSNR | Mixed |
Closest papers by meaning
- APR-RD: Complemental Two Steps for Self-Supervised Real Image Denoising (AAAI 2025 · similarity 0.91)
- Exploring Efficient Asymmetric Blind-Spots for Self-Supervised Denoising in Real-World Scenarios (CVPR 2024 · similarity 0.91)
- Pseudo-Siamese Blind-spot Transformers for Self-Supervised Real-World Denoising (NeurIPS 2024 · similarity 0.88)
- Zero-Shot Blind-spot Image Denoising via Implicit Neural Sampling (CVPR 2025 · similarity 0.87)
- Rethinking Transformer-Based Blind-Spot Network for Self-Supervised Image Denoising (AAAI 2025 · similarity 0.87)
- Masked and Shuffled Blind Spot Denoising for Real-World Images (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.