Research map / Image generation / Image restoration
Clean Images are Hard to Reblur: Exploiting the Ill-Posed Inverse Task for Dynamic Scene Deblurring
Read the paperCheck 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 supervised reblurring loss with 2 ResBlocks to U-Net | U-Net · GOPRO | LPIPS: 0.1635 -> 0.1238 | Helped |
| Add supervised reblurring loss with 2 ResBlocks to SRN | SRN · GOPRO | NIQE: 5.252 -> 4.887 | Helped |
| Add supervised reblurring loss with 2 ResBlocks to DHN | DHN · REDS | LPIPS: 0.0942 -> 0.0805 | Helped |
| Add supervised reblurring loss with 2 ResBlocks to U-Net | U-Net · GOPRO | PSNR: 29.66 -> 29.44 | Hurt |
| Add 5-step self-supervised test-time adaptation (ablation) | SRN + reblur n2 · GOPRO | LPIPS: 0.1037 -> 0.0983 | Helped |
| Add 5-step self-supervised test-time adaptation (ablation) | SRN + reblur n2 · GOPRO | PSNR: 30.57 -> 30.44 | Hurt |
| Increase test-time adaptation from 5 to 30 steps (ablation) | SRN + reblur n2 · GOPRO | LPIPS: 0.0983 -> 0.1199 | Hurt |
| Increase test-time adaptation from 5 to 30 steps (ablation) | SRN + reblur n2 · GOPRO | NIQE: 4.730 -> 4.045 | Helped |
| Add sharpness-preservation loss to blur-reconstruction training (ablation) | U-Net + reblurring module · GOPRO | LPIPS: 0.1301 -> 0.1238 | Helped |
| Use real sharp S instead of pseudo-sharp image in sharpness-preservation loss (ablation) | U-Net + reblurring module · GOPRO | LPIPS: 0.1238 -> 0.1410 | Hurt |
| Use 1 rather than 2 ResBlocks in reblurring module (ablation) | U-Net + supervised reblurring loss · GOPRO | LPIPS: 0.1238 -> 0.1365 | Hurt |
| Use 3 rather than 2 ResBlocks in reblurring module (ablation) | U-Net + supervised reblurring loss · GOPRO | LPIPS: 0.1238 -> 0.1386 | Hurt |
Closest papers by meaning
- Blur2Blur: Blur Conversion for Unsupervised Image Deblurring on Unknown Domains (CVPR 2024 · similarity 0.89)
- Parameterized Blur Kernel Prior Learning for Local Motion Deblurring (CVPR 2025 · similarity 0.88)
- Adaptive Window Pruning for Efficient Local Motion Deblurring (ICLR 2024 · similarity 0.88)
- ID-Blau: Image Deblurring by Implicit Diffusion-based reBLurring AUgmentation (CVPR 2024 · similarity 0.88)
- BlurDM: A Blur Diffusion Model for Image Deblurring (NeurIPS 2025 · similarity 0.87)
- GS-Blur: A 3D Scene-Based Dataset for Realistic Image Deblurring (NeurIPS 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.