Research map / Image generation / Image restoration
Denoising Prior Driven Deep Neural Network for Image Restoration
Weisheng Dong, Peiyao Wang, Wotao Yin, Guangming Shi, Fangfang Wu, Xiaotong Lu
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 |
|---|---|---|---|
| add degradation-model (back-projection) layers to the denoising network (ablation) | DPDNN · Set10 deblurring (motion kernel 2, sigma_n=2.55) | PSNR (dB): 32.22 -> 32.64 | Improved |
| add degradation-model layers to the denoising network (ablation) | DPDNN · Urban100 SR (bicubic x3) | PSNR (dB): 26.98 -> 27.61 | Improved |
| add degradation-model layers to the denoising network (ablation) | DPDNN · Set10 deblurring and SR | PSNR (dB): up to 0.42 dB and 0.63 dB | Improved |
| initialize degradation layers with matrix A instead of random (ablation) | DPDNN · Set10 deblurring (Kernel 1, noise 2.55) | PSNR (dB): 33.19 -> 33.24 | No clear effect |
| initialize degradation layers with matrix A instead of random (ablation) | DPDNN · Set14 SR (Gaussian x3) | PSNR (dB): 29.88 -> 29.91 | No clear effect |
| pretrain the denoiser instead of random init (ablation) | DPDNN · Set12 denoising (sigma=15) | PSNR (dB): 32.91 -> 32.86 | No clear effect |
| pretrain the denoiser instead of random init (ablation) | DPDNN · BSD68 denoising (sigma=50) | PSNR (dB): 27.02 -> 27.03 | No clear effect |
| pretrain the denoiser instead of random init (ablation) | DPDNN · BSD100 SR (bicubic x3) | PSNR (dB): 29.00 -> 28.91 | No clear effect |
| use DPDNN instead of MemNet | denoising network · BSD68 denoising (sigma=50) | PSNR (dB): 26.37 -> 27.02 | Improved |
| use DPDNN instead of DD-CNN at higher noise motion blur | deblurring · Set10 deblurring (motion kernel 1, sigma_n=7.65) | PSNR (dB): 29.17 -> 29.01 | Hurt |
| use DPDNN instead of MemNet | SR · Set5 SR (bicubic x3) | PSNR (dB): 34.09 -> 33.93 | Hurt |
| use DPDNN instead of MemNet (fewer parameters) | network complexity · BSD68 sigma=50 | run time (s/image): 0.278 -> 0.161 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MemNet — MemNet: A Persistent Memory Network for Image Restoration (ICCV 2017) · better on 10 of 10 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Unsupervised Image Restoration Using Partially Linear Denoisers (TPAMI 2021 · similarity 0.92)
- Plug-and-Play Image Restoration With Deep Denoiser Prior (TPAMI 2021 · similarity 0.92)
- Revitalizing Convolutional Network for Image Restoration (TPAMI 2024 · similarity 0.91)
- Universal Denoising Networks : A Novel CNN Architecture for Image Denoising (CVPR 2018 · similarity 0.91)
- A Restoration Network as an Implicit Prior (ICLR 2024 · similarity 0.90)
- Image Restoration by Denoising Diffusion Models with Iteratively Preconditioned Guidance (CVPR 2024 · similarity 0.90)
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.
