Research map / Image generation / Image restoration
Deblurring Using Analysis-Synthesis Networks Pair
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 |
|---|---|---|---|
| Add the analysis and synthesis network pair instead of using a synthesis network without analysis (ablation) | Deblurring network · 1000 images in Figure 1 | PSNR: 23.7 -> 26.6 | Helped |
| Provide the synthesis network with the ground-truth blur kernel (ablation) | Synthesis network · 1000 images in Figure 1 | PSNR: 26.6 -> 28.28 | Helped |
| Add kernel-guided convolution to the synthesis network (ablation) | Synthesis network · Validation set | PSNR: 24.80 -> 28.73 | Helped |
| Use additive kernel guidance only (ablation) | Synthesis network · Validation set | PSNR: 24.80 -> 28.58 | Helped |
| Use multiplicative kernel guidance only (ablation) | Synthesis network · Validation set | PSNR: 24.80 -> 28.41 | Helped |
| Use both additive and multiplicative kernel guidance rather than additive guidance only (ablation) | Synthesis network · Validation set | PSNR: 28.58 -> 28.73 | Helped |
| Use both additive and multiplicative kernel guidance rather than multiplicative guidance only (ablation) | Synthesis network · Validation set | PSNR: 28.41 -> 28.73 | Helped |
| Pre-train the analysis and synthesis networks before end-to-end training rather than use random initialization (ablation) | Analysis-synthesis network pair · Validation set | PSNR: 25.70 -> 24.82 | Hurt |
| Use the reported pre-training-after-E2E training strategy rather than random initialization (ablation) | Analysis-synthesis network pair · Validation set | PSNR: 25.70 -> 26.78 | Helped |
| Specialize networks by blur-kernel size range (ablation) | Analysis-synthesis network pair · Validation set | PSNR: 26.78 -> 27.12 | Helped |
| Specialize networks by blur-kernel size range (ablation) | Analysis-synthesis network pair · Köhler dataset, All Trajectories | PSNR: 29.97 -> 30.22 | Helped |
| Specialize networks by blur-kernel size range (ablation) | Analysis-synthesis network pair · Köhler dataset, Excluding 8,9,10 | PSNR: 31.87 -> 32.24 | Helped |
Closest papers by meaning
- Non-Blind Deblurring: Handling Kernel Uncertainty With CNNs (CVPR 2018 · similarity 0.91)
- DeblurDiff: Real-Word Image Deblurring with Generative Diffusion Models (NeurIPS 2025 · similarity 0.89)
- Cross-Scale Self-Supervised Blind Image Deblurring via Implicit Neural Representation (NeurIPS 2024 · similarity 0.88)
- BlurDM: A Blur Diffusion Model for Image Deblurring (NeurIPS 2025 · similarity 0.88)
- Blur2Blur: Blur Conversion for Unsupervised Image Deblurring on Unknown Domains (CVPR 2024 · similarity 0.88)
- Unsupervised Blind Image Deblurring Based on Self-Enhancement (CVPR 2024 · similarity 0.88)
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.