Research map / Image generation / Image restoration
Learning to Deblur Using Light Field Generated and Real Defocus Images
Lingyan Ruan, Bin Chen, Jizhou Li, Miuling Lam
Check 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 |
|---|---|---|---|
| Train on LFDOF, then fine-tune on DPDD (ablation) | Proposed deblurring network · RealDOF | PSNR: 23.076 -> 25.745; 24.700 -> 25.745 | Improved |
| Apply the LFDOF-to-DPDD training scheme (ablation) | DPDNet_S, KPAC, and proposed network · DPDD | PSNR: +0.511dB, +0.249dB, and +0.253dB | Improved |
| Use one-stage mixed-dataset training instead of two-stage training (ablation) | Proposed deblurring network · RealDOF | PSNR: 24.634 -> 25.745; 24.684 -> 25.745 | Hurt |
| Use L1 rather than VGG feature loss for DPDD fine-tuning (ablation) | Proposed deblurring network · RealDOF | PSNR: 25.130 vs 25.745; 0.62dB lower | Hurt |
| Add the direct input-image component to the dynamic residual block (ablation) | Dynamic residual block · RealDOF | PSNR: 25.576 -> 25.745 | Improved |
| Add the learned dynamic residual component (ablation) | Dynamic residual block · RealDOF | PSNR: 25.515 -> 25.745 | Improved |
| Add the direct residual component (ablation) | Dynamic residual block · RealDOF | PSNR: 25.539 -> 25.745 | Improved |
| Use multi-scale restoration (ablation) | Dynamic residual block network · RealDOF | PSNR: 25.532 -> 25.745 | Improved |
| Set dynamic kernel size to 9 instead of 7 (ablation) | Proposed network · DPDD | PSNR: 25.725 -> 25.752 | Improved |
| Set dynamic kernel size to 9 instead of 7 (ablation) | Proposed network · RealDOF | PSNR: 25.745 -> 25.552 | Hurt |
| Use the end-to-end architecture without explicit defocus-map estimation | Proposed network versus AIFNet · LFDOF | PSNR: +0.726dB | Improved |
Closest papers by meaning
- Defocus Deblurring Using Dual-Pixel Data (ECCV 2020 · similarity 0.94)
- Learning To Reduce Defocus Blur by Realistically Modeling Dual-Pixel Data (ICCV 2021 · similarity 0.92)
- Learnable Blur Kernel for Single-Image Defocus Deblurring in the Wild (AAAI 2023 · similarity 0.92)
- Residual Diffusion Deblurring Model for Single Image Defocus Deblurring (AAAI 2025 · similarity 0.91)
- Quad-Pixel Image Defocus Deblurring: A New Benchmark and Model (CVPR 2025 · similarity 0.91)
- Prior and Prediction Inverse Kernel Transformer for Single Image Defocus Deblurring (AAAI 2024 · similarity 0.90)
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.
