Research map / Image generation / Image restoration
Self-Supervised Non-Uniform Kernel Estimation With Flow-Based Motion Prior for Blind Image 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 |
|---|---|---|---|
| Use flow-based latent motion priors for non-uniform kernel estimation (ablation) | KE-Net · GoPro | PSNR (reblur): 43.90 -> 44.56 | Helped |
| Add uncertainty learning to flow-prior kernel estimation (ablation) | KE-Net · GoPro | PSNR (reblur): 44.56 -> 45.92 | Helped |
| Add flow-prior kernel estimation and uncertainty learning to NAFNet (ablation) | NAFNet · GoPro, HIDE, RealBlur-R, RealBlur-J | PSNR: 33.69 -> 34.06; 31.32 -> 31.74; 35.50 -> 36.25; 28.32 -> 29.87 | Helped |
| Use Whyte et al. kernel estimation instead of omitting kernel estimation (ablation) | NAFNet · GoPro, HIDE, RealBlur-R, RealBlur-J | PSNR: 33.69 -> 33.74; 31.32 -> 31.38; 35.50 -> 35.61; 28.32 -> 28.79 | Helped |
| Add the proposed kernel estimation module, including kernel estimation and multi-scale kernel attention (ablation) | MIMO-UNet, MPRNet, and NAFNet · GoPro and HIDE | PSNR: MIMO-UNet: 32.45 -> 32.83, 29.99 -> 30.16; MPRNet: 32.66 -> 33.04, 30.96 -> 31.13; NAFNet: 33.69 -> 34.06, 31.32 -> 31.7 | Helped |
| Use UFPNet instead of the strongest listed baseline | Blind image deblurring models trained on GoPro · GoPro, HIDE, RealBlur-R, and RealBlur-J | PSNR: 33.69 -> 34.06; 31.42 -> 31.74; 36.07 -> 36.25; 28.97 -> 29.87 | Helped |
| Use UFPNet instead of the strongest listed baseline | Models trained and tested on RealBlur · RealBlur-R and RealBlur-J | PSNR: 39.84 -> 40.61; 32.48 -> 33.35 | Helped |
Closest papers by meaning
- Parameterized Blur Kernel Prior Learning for Local Motion Deblurring (CVPR 2025 · similarity 0.91)
- Motion-adaptive Separable Collaborative Filters for Blind Motion Deblurring (CVPR 2024 · similarity 0.88)
- DeblurDiff: Real-Word Image Deblurring with Generative Diffusion Models (NeurIPS 2025 · similarity 0.88)
- Cross-Scale Self-Supervised Blind Image Deblurring via Implicit Neural Representation (NeurIPS 2024 · similarity 0.87)
- Real-World Efficient Blind Motion Deblurring via Blur Pixel Discretization (CVPR 2024 · similarity 0.86)
- Prior and Prediction Inverse Kernel Transformer for Single Image Defocus Deblurring (AAAI 2024 · similarity 0.86)
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.