AI Paper Atlas

Research map / Image generation / Image restoration

Learning Continuous Wasserstein Barycenter Space for Generalized All-in-One Image Restoration

Xiaole Tang, Xiaoyi He, Jiayi Xu, Xiang Gu, Jiayu Sun

TPAMI 2026 · Image generation › Image restoration · approach: Efficient restoration architectures

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 Wasserstein barycenter learning with PromptIR backbonePromptIR · 3 degradations averagePSNR: 32.05 -> 33.13Improved
add Wasserstein barycenter learning with Restormer backboneRestormer · 5 degradations averagePSNR: 27.46 -> 30.72Improved
add Wasserstein barycenter learning with PromptIR backbonePromptIR · 5 degradations averagePSNR: 29.72 -> 31.05Improved
use WB embeddings instead of original embeddings (ablation)BaryIR · 5 degradations averagePSNR: 27.46 -> 29.89Improved
add residual embeddings to WB embeddings (ablation)BaryIR · 5 degradations averagePSNR: 29.89 -> 30.72Improved
add residual embeddings to original embeddings (no WB) (ablation)Restormer · O-HAZEPSNR: 18.02 -> 19.84Improved
add all three losses MWB, IRC, BRO (ablation)BaryIR · 5 degradations averagePSNR: 27.72 -> 30.72Improved
add MWB loss alone (ablation)BaryIR · 5 degradations averagePSNR: 27.72 -> 29.37Improved
use transformer-block transport map instead of MLP (ablation)BaryIR · 5 degradations averagePSNR: 30.04 -> 30.72Improved
use portion-based barycenter weights instead of uniform (ablation)BaryIR · 3 degradations averagePSNR: 32.67 -> 32.86Improved
increase training batch size beyond 8 (ablation)BaryIR · 5 degradationsPSNRNo clear effect
apply BaryIR to unseen JPEG artifactsBaryIR vs MoCE-IR · BSD500 QF=10PSNR: 26.42 -> 27.94Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 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