AI Paper Atlas

Research map / Image generation / Image restoration

Deep Image Deblurring: A Survey

Kaihao Zhang, Wenqi Ren, Wenhan Luo, Wei‐Sheng Lai, Björn Stenger, Ming–Hsuan Yang, Hongdong Li

IJCV 2022 · Image generation › Image restoration · approach: Image deblurring · 441 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
add perceptual and GAN loss to L2 loss (ablation)single image deblurring network · Single image deblurringPSNR: 26.78 -> 27.20Improved
add perceptual and GAN loss to L1 loss (ablation)single image deblurring network · Single image deblurringPSNR: 26.24 -> 26.89Improved
use RaGAN instead of GAN with L2 + perceptual (ablation)single image deblurring network · Single image deblurringPSNR: 27.20 -> 27.09Hurt
use L2 instead of L1 pixel loss (ablation)single image deblurring network · Single image deblurringPSNR: 26.24 -> 26.78Improved
train on RealBlur-J + BSD-B + GoPro instead of RealBlur-J only (ablation)SRN · RealBlur-J testPSNR: 31.02 -> 31.37Improved
train on GoPro only instead of in-domain RealBlur-J (ablation)SRN · RealBlur-J testPSNR: 31.02 -> 28.56Hurt
provide ground-truth blur kernels (non-blind instead of blind)USRNet vs SRN · Non-blind GoProPSNR: 34.78 -> 36.51Improved
use multi-scale architecture for UHD deblurringSRN vs DeblurGAN · UHD image deblurringPSNR: 19.25 -> 21.25Improved
use video deblurring network instead of single image networkPan et al. vs Tao et al. · DVDPSNR: 29.97 -> 32.13Improved
use multi-scale scale-recurrent net instead of conditional GANTao et al. vs Kupyn et al. · GoProPSNR: 28.70 -> 30.26Improved

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