Research map / Image generation / Image restoration
When AWGN-Based Denoiser Meets Real Noises
Yuqian Zhou, Jianbo Jiao, Haibin Huang, Yang Wang, Jue Wang, Honghui Shi, Thomas Huang
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 |
|---|---|---|---|
| Train with AWGN-RVIN and explicitly estimate the two noise types | Proposed denoiser · Set20 AWGN-RVIN | PSNR: 32.09 -> 32.43; 29.97 -> 30.47; 29.52 -> 29.82; 27.90 -> 28.41 | Improved |
| Use the proposed blind model instead of DnCNN-B | Signal-dependent noise denoising · CBSD68 (20, 10) | PSNR: 34.38 -> 34.75 | Improved |
| Use the proposed blind model instead of DnCNN-B | Signal-dependent noise denoising · CBSD68 (20, 20) | PSNR: 31.72 -> 31.32 | Hurt |
| Use the proposed blind model instead of DnCNN-B | Signal-dependent noise denoising · CBSD68 (40, 10) and (40, 20) | PSNR: 32.08 -> 32.12; 30.32 -> 30.33 | Mixed |
| Apply PD adaptation to the AWGN-RVIN base model | Proposed denoiser · DND | PSNR: 32.60 -> 38.40 | Improved |
| Apply PD adaptation to existing denoisers | CBM3D, CDnCNN-B, and FFDNet · DND | PSNR: 34.51 -> 35.02; 32.43 -> 35.44; 34.40 -> 37.56 | Improved |
| Add RVIN to AWGN training (ablation) | Proposed denoiser · Real noisy night photos | dead-pixel removal | Improved |
| Use stride s=1 rather than the full pipeline's s=2 (ablation) | PD refinement · DND | PSNR: 38.40 -> 32.60 | Hurt |
| Use stride s=3 rather than the full pipeline's s=2 (ablation) | PD refinement · DND | PSNR: 38.40 -> 37.90 | Hurt |
| Stop after pixel-shuffle upsampling (I), omitting later refinement (ablation) | PD refinement · DND | PSNR: 38.40 -> 37.00 | Hurt |
| Use DI refinement instead of the full pipeline (ablation) | PD refinement · DND | PSNR: 38.40 -> 37.20 | Hurt |
| Increase blending factor k (ablation) | PD refinement · Real noisy images | detail preservation | Hurt |
Closest papers by meaning
- sRGB Real Noise Synthesizing With Neighboring Correlation-Aware Noise Model (CVPR 2023 · similarity 0.90)
- Toward Convolutional Blind Denoising of Real Photographs (CVPR 2019 · similarity 0.89)
- An Unsupervised Deep Learning Approach for Real-World Image Denoising (ICLR 2021 · similarity 0.89)
- APR-RD: Complemental Two Steps for Self-Supervised Real Image Denoising (AAAI 2025 · similarity 0.88)
- Calibration-Free Raw Image Denoising via Fine-Grained Noise Estimation (TPAMI 2025 · similarity 0.88)
- sRGB Real Noise Modeling via Noise-Aware Sampling with Normalizing Flows (ICLR 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.
