All-in-One Image Restoration

Overview

All-in-one image restoration studies how one model can recover images affected by different problems—such as rain, haze, blur, noise, or combinations of them—without needing a separate system for each. Common approaches use degradation-aware prompts or physical priors, specialized experts with routing, and shared CNN or transformer backbones; diffusion and other generative models are also used to reconstruct missing detail. Researchers test these systems on standard task benchmarks and increasingly on real-world, mixed-degradation, and unseen conditions, sometimes measuring downstream performance as well as image quality. A central challenge is balancing specialization with generalization: models must remove artifacts without erasing meaningful content or inventing details, while remaining efficient and controllable.

Latest papers

CardsList

93 papers5 papers in the last four weeks, down 50% on the four weeks before. 0.0% of all new papers.

  1. Rectification Reimagined: A Unified Mamba Model for Image Correction and Rectangling with Prompts

    Dec 21, 2025Linwei Qiu, Gongzhe Li, Xiaozhe Zhang +2All-in-One Image Restoration

  2. ClusIR: Towards Cluster-Guided All-in-One Image Restoration

    Dec 11, 2025Shengkai Hu, Jiaqi Ma, Xu Zhang +3Mixture of ExpertsAll-in-One Image Restoration