Restora-Flow: Mask-Guided Image Restoration with Flow Matching

TL;DR

Restora-Flow is a training-free, flow-guided image restoration method that improves processing speed and output quality across various tasks by using degradation masks and trajectory correction.

Contribution

It introduces a novel, training-free approach that guides flow matching sampling with degradation masks and enforces input consistency, addressing speed and quality issues in flow-based restoration.

Findings

Outperforms diffusion and flow matching methods in perceptual quality.

Reduces processing time significantly.

Effective across natural and medical image restoration tasks.

Abstract

Flow matching has emerged as a promising generative approach that addresses the lengthy sampling times associated with state-of-the-art diffusion models and enables a more flexible trajectory design, while maintaining high-quality image generation. This capability makes it suitable as a generative prior for image restoration tasks. Although current methods leveraging flow models have shown promising results in restoration, some still suffer from long processing times or produce over-smoothed results. To address these challenges, we introduce Restora-Flow, a training-free method that guides flow matching sampling by a degradation mask and incorporates a trajectory correction mechanism to enforce consistency with degraded inputs. We evaluate our approach on both natural and medical datasets across several image restoration tasks involving a mask-based degradation, i.e., inpainting, super-resolution and denoising. We show superior perceptual quality and processing time compared to diffusion and flow matching-based reference methods.