Research map / Image generation / Image restoration
Learning Fully Convolutional Networks for Iterative Non-Blind Deconvolution
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 three iterative stages instead of one (ablation) | FCNN deconvolution · Sun/Levin benchmark, 1% noise, ground-truth kernel | PSNR/SSIM: 32.50/0.89 -> 32.82/0.90 | Mixed |
| Use three iterative stages instead of one (ablation) | FCNN deconvolution · Sun/Levin benchmark, 1% noise, Pan kernel | PSNR/SSIM: 30.38/0.86 -> 30.39/0.87 | Mixed |
| Use three iterative stages instead of one (ablation) | FCNN deconvolution · Sun/Levin benchmark, 3% noise, ground-truth kernel | PSNR/SSIM: 28.71/0.77 -> 29.12/0.79 | Mixed |
| Use three iterative stages instead of one (ablation) | FCNN deconvolution · Sun/Levin benchmark, 3% noise, Pan kernel | PSNR/SSIM: 27.34/0.75 -> 27.58/0.77 | Helped |
| Use three iterative stages instead of one (ablation) | FCNN deconvolution · Sun/Levin benchmark, 5% noise, ground-truth kernel | PSNR/SSIM: 27.25/0.72 -> 27.66/0.74 | Helped |
| Use three iterative stages instead of one (ablation) | FCNN deconvolution · Sun/Levin benchmark, 5% noise, Pan kernel | PSNR/SSIM: 25.49/0.69 -> 25.77/0.72 | Mixed |
| Train the FCNN in the gradient domain rather than the intensity domain (ablation) | FCNN deconvolution · Visual comparison across iterative deconvolution | noise and ringing artifacts | Helped |
| Use an L1 reconstruction loss instead of L2 (ablation) | FCNN gradient denoiser · First iteration, images with 1% noise | gradient denoising quality | Helped |
| Use three iterations instead of one (ablation) | FCNN deconvolution · Image deconvolution, qualitative analysis | PSNR | Helped |
Closest papers by meaning
- Neural Blind Deconvolution Using Deep Priors (CVPR 2020 · similarity 0.91)
- Deep Non-Blind Deconvolution via Generalized Low-Rank Approximation (NeurIPS 2018 · similarity 0.90)
- Universal Denoising Networks : A Novel CNN Architecture for Image Denoising (CVPR 2018 · similarity 0.90)
- Robust And Interpretable Blind Image Denoising Via Bias-Free Convolutional Neural Networks (ICLR 2020 · similarity 0.89)
- Deep Learning for Handling Kernel/model Uncertainty in Image Deconvolution (CVPR 2020 · similarity 0.89)
- Network Deconvolution (ICLR 2020 · similarity 0.89)
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.