Research map / Image generation / Image restoration
Universal Denoising Networks : A Novel CNN Architecture 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 |
|---|---|---|---|
| Use non-local filtering instead of local convolutional operators (ablation) | UNet₅ · 68 Berkeley images, grayscale AWGN, σ=5–55 average | PSNR (dB): 29.22 -> 29.32 | Helped |
| Use non-local filtering instead of local convolutional operators (ablation) | CUNet₅ · 68 Berkeley images, color AWGN, σ=5–55 average | PSNR (dB): 31.35 -> 31.49 | Helped |
| Use ground-truth images to compute similar-patch indices (ablation) | UNLNet₅ · 68 Berkeley images, grayscale AWGN, σ=5–55 average | PSNR (dB): +0.25 dB | Helped |
| Use ground-truth images to compute similar-patch indices (ablation) | CUNLNet₅ · 68 Berkeley images, color AWGN, σ=5–55 average | PSNR (dB): 31.49 -> 32.18 | Helped |
| Use UNLNet₅ instead of UNet₅ | Proposed grayscale denoising networks · 68 Berkeley images, grayscale AWGN, σ=5–55 average | PSNR (dB): UNLNet₅ performs better than the local one but still falls behind DnCNN around 0.15 dB on average. | Hurt |
| Use UNet₅ instead of DnCNN | Grayscale denoising · 68 Berkeley images, grayscale AWGN, σ=5–55 average | PSNR (dB): 0.25 dB worse | Hurt |
| Use CUNLNet₅ instead of CDnCNN | Color denoising · 68 Berkeley images, color AWGN, σ=5–55 average | PSNR (dB): 31.49 vs. 31.48 | No clear effect |
| Use proposed color networks instead of CBM3D | Color denoising · 68 Berkeley images, color AWGN, σ=5–55 average | PSNR (dB): 0.35 dB for the local model; 0.5 dB for the non-local model | Helped |
| Train a single model across a range of noise levels using the constrained formulation | Proposed networks · Grayscale and color denoising across noise levels | PSNR | Helped |
Closest papers by meaning
- Non-Local Color Image Denoising With Convolutional Neural Networks (CVPR 2017 · similarity 0.93)
- Robust And Interpretable Blind Image Denoising Via Bias-Free Convolutional Neural Networks (ICLR 2020 · similarity 0.93)
- Self-Supervised Image Denoising Using Implicit Deep Denoiser Prior (AAAI 2023 · similarity 0.92)
- Toward Convolutional Blind Denoising of Real Photographs (CVPR 2019 · similarity 0.92)
- High-Quality Self-Supervised Deep Image Denoising (NeurIPS 2019 · similarity 0.91)
- IDR: Self-Supervised Image Denoising via Iterative Data Refinement (CVPR 2022 · similarity 0.91)
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.