AI Paper Atlas

Research map / Image generation / Super-resolution

A New Dataset and Framework for Real-World Blurred Images Super-Resolution

Rui Qin, Ming Sun, Chao Zhou, Bin Wang

ECCV 2024 · Image generation › Super-resolution · approach: Blind super-resolution · 1 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
train unified on mixed blur and general data instead of separate data types (ablation)FeMaSR · ReBlurSR-Test and general dataLPIPSHurt
average two CDM branches after training without CFM (ablation)CDM_FeMaSR · ReBlurSR-Test (blur data)LPIPS: 0.3714 -> 0.3711No clear effect
add cross fusion module (CFM) during CDM training (ablation)PBaSR · general and blur dataLPIPS: 0.01-0.015 LPIPSImproved
condition discriminator on blur map in blur branch (ablation)FeMaSR blur branch · ReBlurSR-Test HRLPIPS on blur samples: 0.022 LPIPS improvementImproved
add synthetic blur images to ReBlurSR training data (ablation)FeMaSR · ReBlurSR-Test (blur data)LPIPS: about 0.013 LPIPSImproved
CFM interpolation communication every iteration (k=1) or every 100 (k=100) instead of k=20 (ablation)PBaSR_FeMaSR · ReBlurSR-Test (Defocus)LPIPS: 0.005-0.01 LPIPSHurt
use lambda0 between 0.9 and 0.99 instead of 0 or 1 (ablation)PBaSR_FeMaSR · ReBlurSR-Test (Defocus)LPIPSImproved
use CFM weight interpolation instead of cross feature distillation or dual teacher-student (ablation)PBaSR · blur and general dataLPIPS: 0.006 to 0.008 LPIPSImproved
apply PBaSR to FeMaSR vs best prior on blur data (HAT)PBaSR_FeMaSR · ReBlurSR-TestLPIPS: 0.3924 -> 0.3564Improved
apply PBaSR to Real-ESRGAN vs gan artifact correction methods LDL and DeSRAReal-ESRGAN · ReBlurSR-Test and generalLPIPS: 0.007-0.02 LPIPS on blur dataImproved
add defocus blur training data progressively (ablation)PBaSR · ReBlurSR-Test (defocus)LPIPS: 0.04 LPIPSImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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