Research map / Image generation / Image restoration
Learning Deep CNN Denoiser Prior for Image Restoration
Check 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 the CNN denoiser as the prior in HQS deblurring | HQS image deblurring · Leaves, Gaussian blur σ=1.6, noise σ=2 | PSNR (dB): 28.91 -> 29.78 | Helped |
| Use a CNN denoiser prior for color denoising | BSD68 color denoising · BSD68 color, σ=25 | PSNR (dB): 30.71 -> 31.16 | Helped |
| Use a CNN denoiser prior for Gaussian-kernel super-resolution | Single-image super-resolution · Set5, scale 3, Gaussian kernel, Y channel | PSNR (dB): 33.02 -> 33.38 | Helped |
| Use dilated convolutions instead of standard 3×3 convolutions at 7 layers (ablation) | CNN denoiser · BSD68, σ=25 | PSNR (dB): 28.94 -> 29.15 | Helped |
| Use dilated convolutions instead of standard 3×3 convolutions at the same receptive field (ablation) | CNN denoiser · BSD68, σ=25 | PSNR (dB): 29.20 -> 29.15 | No clear effect |
| Train using larger image patches instead of small patches (ablation) | CNN denoiser · Image denoising | denoising performance | No clear effect |
| Use training patches smaller than the network receptive field (ablation) | CNN denoiser · Image denoising | denoising performance | Hurt |
| Increase training data from BSD images to the larger mixed image dataset (ablation) | CNN denoiser · BSD68 and other testing images | PSNR | Mixed |
| Reduce middle-layer feature maps from 64 to 24 (ablation) | CNN denoiser · Image denoising, noise level 25 | PSNR (dB) | No clear effect |
Closest papers by meaning
- Universal Denoising Networks : A Novel CNN Architecture for Image Denoising (CVPR 2018 · similarity 0.90)
- A Restoration Network as an Implicit Prior (ICLR 2024 · similarity 0.89)
- Stochastic Solutions for Linear Inverse Problems using the Prior Implicit in a Denoiser (NeurIPS 2021 · similarity 0.89)
- Denoising as Adaptation: Noise-Space Domain Adaptation for Image Restoration (ICLR 2025 · similarity 0.88)
- Image Restoration by Denoising Diffusion Models with Iteratively Preconditioned Guidance (CVPR 2024 · similarity 0.88)
- Unleashing the Denoising Capability of Diffusion Prior for Solving Inverse Problems (NeurIPS 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.