Research map / Image generation / Image restoration
Douglas-Rachford Networks: Learning Both the Image Prior and Data Fidelity Terms for Blind Image Deconvolution
Raied Aljadaany, Dipan K. Pal, Marios Savvides
Check 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 |
|---|---|---|---|
| Learn both the image-prior and data-fidelity proximal operators instead of replacing either with a hand-crafted or estimated-kernel term (ablation) | Dr-Net · Ablation test set | PSNR and SSIM | Improved |
| Add non-expansive constraints to the convolutional filters (ablation) | Dr-Net · GoPro test dataset | PSNR and SSIM | Improved |
| Increase the number of Dr-Blocks from 3 to 5 (ablation) | Dr-Net · Ablation test set | PSNR (dB): 27.26 -> 30.40 | Improved |
| Increase the number of Dr-Blocks from 5 to 7 (ablation) | Dr-Net · Ablation test set | PSNR (dB): 30.40 -> 30.71 | Improved |
| Increase the number of Dr-Blocks from 7 to 9 (ablation) | Dr-Net · Ablation test set | PSNR (dB): 30.71 -> 30.93 | Improved |
| Increase the number of Dr-Blocks from 3 to 5 (ablation) | Dr-Net · Ablation test set | Time (sec): 49.1 -> 57.4 | Hurt |
| Add the pre-processing network (ablation) | Dr-Net · Ablation test set | PSNR (dB): 28.57 -> 30.40 | Improved |
| Add the post-processing network (ablation) | Dr-Net · Ablation test set | PSNR (dB): 28.20 -> 30.40 | Improved |
| Add both the pre-processing and post-processing networks (ablation) | Dr-Net · Ablation test set | PSNR (dB): 27.12 -> 30.40 | Improved |
| Use Dr-Net rather than Tao | Blind deconvolution · GoPro test dataset | PSNR (dB): 30.26 -> 30.35 | Improved |
| Use Dr-Net rather than Xu | Blind deconvolution · Kohler dataset | PSNR (dB): 27.47 -> 27.20 | Hurt |
| Use Dr-Net rather than Tao | Blind deconvolution · Kohler dataset | SSIM: 0.837 -> 0.865 | Improved |
Closest papers by meaning
- Neural Blind Deconvolution Using Deep Priors (CVPR 2020 · similarity 0.89)
- Learning to Deblur (TPAMI 2016 · similarity 0.89)
- Deblurring Using Analysis-Synthesis Networks Pair (CVPR 2020 · similarity 0.88)
- Training Image Estimators without Image Ground Truth (NeurIPS 2019 · similarity 0.88)
- DeepSN-Net: Deep Semi-Smooth Newton Driven Network for Blind Image Restoration (TPAMI 2025 · similarity 0.87)
- A Restoration Network as an Implicit Prior (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.
