Research map / Image generation / Image restoration
Real-World Blur Dataset for Learning and Benchmarking Deblurring Algorithms
Jaesung Rim, Haeyun Lee, Jucheol Won, Sunghyun Cho
Read the paperCheck 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 |
|---|---|---|---|
| add blur-kernel-based alignment to reference homography (ablation) | dataset alignment · RealBlur alignment | displacement error (pixels): 4.9454 -> 0.8701 | Improved |
| add phase correlation to homography plus blur kernel alignment (ablation) | dataset alignment · RealBlur alignment | displacement error (pixels): 0.8701 -> 0.8058 | Improved |
| align data with blur-kernel centroids instead of phase correlation (ablation) | SRN-DeblurNet · Kohler et al. test | PSNR (dB): 26.48 -> 28.07 | Improved |
| align data with centroids instead of non-zero centers (ablation) | SRN-DeblurNet · Kohler et al. test | PSNR (dB): 27.80 -> 28.07 | Improved |
| train on RealBlur-J instead of GoPro (ablation) | SRN-DeblurNet · RealBlur-J test | PSNR (dB): 28.56 -> 31.02 | Improved |
| train on RealBlur-R instead of GoPro (ablation) | SRN-DeblurNet · RealBlur-R test | PSNR (dB): 35.66 -> 36.47 | Improved |
| train on BSD-B instead of GoPro (ablation) | SRN-DeblurNet · RealBlur-R test | PSNR (dB): 35.66 -> 34.96 | Hurt |
| combine all training sets with pre-trained weights versus RealBlur-R only (ablation) | SRN-DeblurNet · RealBlur-R test | PSNR (dB): 36.47 -> 38.65 | Improved |
| combine all training sets with pre-trained weights versus RealBlur-J only (ablation) | SRN-DeblurNet · RealBlur-J test | PSNR (dB): 31.02 -> 31.38 | Improved |
| train SRN-DeblurNet on RealBlur data instead of released pretrained model | SRN-DeblurNet · RealBlur-J test | PSNR (dB): 28.56 -> 31.38 | Improved |
| train SRN-DeblurNet on RealBlur data instead of released pretrained model | SRN-DeblurNet · RealBlur-R test | PSNR (dB): 35.66 -> 38.65 | Improved |
| train DeblurGAN-v2 on RealBlur data instead of released pretrained model | DeblurGAN-v2 · RealBlur-J test | PSNR (dB): 28.70 -> 29.69 | Improved |
Closest papers by meaning
- A Comparative Study for Single Image Blind Deblurring (CVPR 2016 · similarity 0.92)
- GS-Blur: A 3D Scene-Based Dataset for Realistic Image Deblurring (NeurIPS 2024 · similarity 0.91)
- Real-World Efficient Blind Motion Deblurring via Blur Pixel Discretization (CVPR 2024 · similarity 0.90)
- Blur2Blur: Blur Conversion for Unsupervised Image Deblurring on Unknown Domains (CVPR 2024 · similarity 0.90)
- Unsupervised Blind Image Deblurring Based on Self-Enhancement (CVPR 2024 · similarity 0.88)
- Adaptive Window Pruning for Efficient Local Motion Deblurring (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.
