AI Paper Atlas

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

CVPR 2022 · Image generation › Image restoration · approach: Image deblurring · 97 citations (OpenAlex)

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 changedWhereMeasuredEffect
Train on LFDOF, then fine-tune on DPDD (ablation)Proposed deblurring network · RealDOFPSNR: 23.076 -> 25.745; 24.700 -> 25.745Improved
Apply the LFDOF-to-DPDD training scheme (ablation)DPDNet_S, KPAC, and proposed network · DPDDPSNR: +0.511dB, +0.249dB, and +0.253dBImproved
Use one-stage mixed-dataset training instead of two-stage training (ablation)Proposed deblurring network · RealDOFPSNR: 24.634 -> 25.745; 24.684 -> 25.745Hurt
Use L1 rather than VGG feature loss for DPDD fine-tuning (ablation)Proposed deblurring network · RealDOFPSNR: 25.130 vs 25.745; 0.62dB lowerHurt
Add the direct input-image component to the dynamic residual block (ablation)Dynamic residual block · RealDOFPSNR: 25.576 -> 25.745Improved
Add the learned dynamic residual component (ablation)Dynamic residual block · RealDOFPSNR: 25.515 -> 25.745Improved
Add the direct residual component (ablation)Dynamic residual block · RealDOFPSNR: 25.539 -> 25.745Improved
Use multi-scale restoration (ablation)Dynamic residual block network · RealDOFPSNR: 25.532 -> 25.745Improved
Set dynamic kernel size to 9 instead of 7 (ablation)Proposed network · DPDDPSNR: 25.725 -> 25.752Improved
Set dynamic kernel size to 9 instead of 7 (ablation)Proposed network · RealDOFPSNR: 25.745 -> 25.552Hurt
Use the end-to-end architecture without explicit defocus-map estimationProposed network versus AIFNet · LFDOFPSNR: +0.726dBImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy