AI Paper Atlas

Research map / Image generation / Image restoration

Learnable Blur Kernel for Single-Image Defocus Deblurring in the Wild

Jucai Zhai, Pengcheng Zeng, Chihao Ma, Jie Chen, Yong Zhao

AAAI 2023 · Image generation › Image restoration · approach: Image deblurring · 6 citations (OpenAlex)

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 changedWhereMeasuredEffect
guide deblurring with learned defocus maps and use DefocusGANsingle-image defocus deblurring · DPDD Indoor & OutdoorPSNR: 25.35 -> 25.56Improved
guide deblurring with learned defocus maps and use DefocusGANsingle-image defocus deblurring · DPDD Indoor & OutdoorLPIPS: 0.225 -> 0.111Improved
use the proposed system rather than DP-view-input methodsDPDNet and RDPDNet · DPDDPSNR: 25.39 -> 25.56Improved
use a learnable blur kernel instead of a Butterworth kernel (ablation)defocus-map estimation · DPDDPSNR: 23.31 -> 23.56Improved
use a learnable blur kernel instead of a Gaussian kernel (ablation)defocus-map estimation · DPDDPSNR: 22.56 -> 23.56Improved
add RCAB (ablation)single-scale base network · DPDDPSNR: 24.73 -> 25.10Improved
add defocus-map guidance (DGB) (ablation)Base + RCAB · DPDDPSNR: 25.10 -> 25.30Improved
add the multi-scale module (ablation)Base + RCAB + DGB · DPDDPSNR: 25.30 -> 25.47Improved
add perceptual loss (ablation)Base + RCAB + DGB + MS · DPDDLPIPS: 0.172 -> 0.115Improved
add GAN training with defocus adversarial loss (ablation)DefocusGAN with perceptual loss · DPDDLPIPS: 0.115 -> 0.111Improved
replace defocus adversarial loss with doubleGAN loss (ablation)DefocusGAN · DPDDPSNR: 25.42 -> 24.92Hurt
replace defocus adversarial loss with doubleGAN loss (ablation)DefocusGAN · DPDDLPIPS: 0.160 -> 0.132Improved

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