Research map / Image generation / Image restoration
Deep Richardson–Lucy Deconvolution for Low-Light Image Deblurring
Liang Chen, Jiawei Zhang, Zhenhua Li, Yunxuan Wei, Faming Fang, Jimmy Ren, Jinshan Pan
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 learnable latent map M (ablation) | Deep RL deconvolution · Saturated test set (GT kernels) | PSNR: 23.34 -> 25.66 | Improved |
| add learnable latent map M (ablation) | Deep RL deconvolution · Saturated test set (estimated kernels) | PSNR: 23.18 -> 25.11 | Improved |
| use CNN latent map instead of Whyte approximation (ablation) | Deep RL deconvolution · Saturated test set (GT kernels) | PSNR: 24.07 -> 25.66 | Improved |
| use CNN latent map instead of NBDN saturated-pixel discarding (ablation) | Deep RL deconvolution · Saturated test set (GT kernels) | PSNR: 24.69 -> 25.66 | Improved |
| use CNN latent map instead of fixed-threshold map (ablation) | Deep RL deconvolution · Saturated test set (GT kernels) | PSNR: 24.13 -> 25.66 | Improved |
| add learned prior PEN (ablation) | Deep RL deconvolution · Saturated test set (GT kernels) | PSNR: 23.88 -> 25.66 | Improved |
| learned PEN prior instead of hyper-Laplacian prior (ablation) | Deep RL deconvolution · Saturated test set (GT kernels) | PSNR: 24.30 -> 25.66 | Improved |
| learned PEN prior instead of hyper-Laplacian prior (ablation) | Deep RL deconvolution · Saturated test set (estimated kernels) | PSNR: 23.79 -> 25.11 | Improved |
| increase iterations from 10 to 30 (ablation) | MEN latent map · Saturated test set (GT kernels) | MSE of M: 0.0123 -> 0.0077 | Improved |
| add PEN prior term in iterations (ablation) | Deep RL deconvolution · Saturated test set (GT kernels) | SSIM: 0.8591 -> 0.8595 | Improved |
| more than 30 iteration stages (ablation) | Deep RL deconvolution · Saturated test set (GT kernels) | PSNR/SSIM | No clear effect |
| synthesize saturation by enlarging only pixels above thresholds instead of whole image (ablation) | Deep RL deconvolution · RealBlur saturated samples | PSNR: 27.89 -> 28.13 | 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.
- DWDN — DWDN: Deep Wiener Deconvolution Network for Non-Blind Image Deblurring (TPAMI 2021) · better on 5 of 5 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
- Learning Spatially-Variant MAP Models for Non-Blind Image Deblurring (CVPR 2021 · similarity 0.89)
- Blind Deblurring for Saturated Images (CVPR 2021 · similarity 0.89)
- Convergence Analysis of MAP Based Blur Kernel Estimation (ICCV 2017 · similarity 0.88)
- Neural Blind Deconvolution Using Deep Priors (CVPR 2020 · similarity 0.88)
- DeblurDiff: Real-Word Image Deblurring with Generative Diffusion Models (NeurIPS 2025 · similarity 0.87)
- Residual Diffusion Deblurring Model for Single Image Defocus Deblurring (AAAI 2025 · 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.
