Research map / Image generation / Image restoration
Realistic Blur Synthesis for Learning Image Deblurring
Jaesung Rim, Geonung Kim, Jungeon Kim, Junyong Lee, Seungyong Lee, 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 |
|---|---|---|---|
| use sRGB gamma CRF instead of linear in naive averaging (ablation) | SRN-DeblurNet on GoPro-style synthesis · RSBlur test | PSNR: 30.12 -> 30.90 | Improved |
| add frame interpolation without noise (ablation) | blur synthesis · RSBlur test | PSNR: 30.90 -> 30.20 | Hurt |
| add Gaussian noise to interpolated blur (ablation) | blur synthesis · RSBlur test | PSNR: 30.20 -> 31.77 | Improved |
| add Gaussian noise to non-interpolated blur (ablation) | blur synthesis · RSBlur test | PSNR: 30.90 -> 31.69 | Improved |
| add oracle saturated pixels (ablation) | blur synthesis · RSBlur test | PSNR: 31.77 -> 31.89 | Improved |
| add our saturation synthesis (ablation) | blur synthesis · RSBlur test | PSNR: 31.77 -> 31.83 | Improved |
| add Poisson noise and ISP on top of Gaussian noise (ablation) | blur synthesis · RSBlur test | PSNR: 31.83 -> 32.06 | Improved |
| our full pipeline versus real training data (ablation) | blur synthesis · RSBlur test | PSNR: 32.06 vs 32.53 | Improved |
| add our saturation synthesis (GoPro) (ablation) | blur synthesis · RealBlur_J | PSNR: 29.17 -> 29.95 | Improved |
| use A7R3 ISP and G+P noise instead of G noise (ablation) | GoPro synthesis · BSD_All | PSNR: 31.41 -> 30.48 | Hurt |
| use convolution-based blur (GoPro_U) with full pipeline (ablation) | blur synthesis · RealBlur_J | PSNR: 29.28 -> 30.75 | Improved |
| use rough instead of accurate camera parameters (ablation) | synthesis pipeline · RealBlur_J | PSNR: 30.32 -> 30.23 | Hurt |
Closest papers by meaning
- GS-Blur: A 3D Scene-Based Dataset for Realistic Image Deblurring (NeurIPS 2024 · similarity 0.91)
- Blur2Blur: Blur Conversion for Unsupervised Image Deblurring on Unknown Domains (CVPR 2024 · similarity 0.90)
- DeblurDiff: Real-Word Image Deblurring with Generative Diffusion Models (NeurIPS 2025 · similarity 0.87)
- ID-Blau: Image Deblurring by Implicit Diffusion-based reBLurring AUgmentation (CVPR 2024 · similarity 0.87)
- BlurDM: A Blur Diffusion Model for Image Deblurring (NeurIPS 2025 · similarity 0.87)
- BLURD: Benchmarking and Learning using a Unified Rendering and Diffusion Model (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.
