UniRes: Universal Image Restoration for Complex Degradations

TL;DR

UniRes introduces a diffusion-based framework that effectively restores images with complex, mixed degradations by leveraging specialized models and a flexible, end-to-end approach, improving performance on real-world degraded images.

Contribution

The paper proposes UniRes, a novel diffusion-based framework that handles complex degradations by combining specialized models, requiring only isolated training data, and allowing flexible extensions and fidelity-quality adjustments.

Findings

Consistent performance improvements on complex degradation datasets.

Effective restoration of in-the-wild images with mixed degradations.

Outperforms existing methods in qualitative and quantitative evaluations.

Abstract

Real-world image restoration is hampered by diverse degradations stemming from varying capture conditions, capture devices and post-processing pipelines. Existing works make improvements through simulating those degradations and leveraging image generative priors, however generalization to in-the-wild data remains an unresolved problem. In this paper, we focus on complex degradations, i.e., arbitrary mixtures of multiple types of known degradations, which is frequently seen in the wild. A simple yet flexible diffusionbased framework, named UniRes, is proposed to address such degradations in an end-to-end manner. It combines several specialized models during the diffusion sampling steps, hence transferring the knowledge from several well-isolated restoration tasks to the restoration of complex in-the-wild degradations. This only requires well-isolated training data for several degradation types. The framework is flexible as extensions can be added through a unified formulation, and the fidelity-quality trade-off can be adjusted through a new paradigm. Our proposed method is evaluated on both complex-degradation and single-degradation image restoration datasets. Extensive qualitative and quantitative experimental results show consistent performance gain especially for images with complex degradations.