MFSR: MeanFlow Distillation for One Step Real-World Image Super Resolution

2026-03-21

AI Summary

This work addresses the limitations of existing single-step image super-resolution methods, which often suffer from quality degradation due to distillation and lose the capacity for multi-step refinement, while multi-step diffusion or flow-based models incur high computational costs during inference. To bridge this gap, we propose the MFSR framework, which— for the first time—adopts MeanFlow as the distillation target, enabling the student model to approximate the average velocity field between arbitrary states of the probability flow ODE. We further enhance detail preservation by refining the Classifier-Free Guidance distillation strategy. MFSR maintains efficient single-step inference while optionally supporting a few iterative refinements for improved quality. Experiments demonstrate that MFSR achieves performance on par with or superior to multi-step teacher models on both synthetic and real-world datasets, substantially reducing computational overhead.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Multimodal Learning
• Multiagent Systems: Multiagent Learning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Diffusion- and flow-based models have advanced Real-world Image Super-Resolution (Real-ISR), but their multi-step sampling makes inference slow and hard to deploy. One-step distillation alleviates the cost, yet often degrades restoration quality and removes the option to refine with more steps. We present Mean Flows for Super-Resolution (MFSR), a new distillation framework that produces photorealistic results in a single step while still allowing an optional few-step path for further improvement. Our approach uses MeanFlow as the learning target, enabling the student to approximate the average velocity between arbitrary states of the Probability Flow ODE (PF-ODE) and effectively capture the teacher's dynamics without explicit rollouts. To better leverage pretrained generative priors, we additionally improve original MeanFlow's Classifier-Free Guidance (CFG) formulation with teacher CFG distillation strategy, which enhances restoration capability and preserves fine details. Experiments on both synthetic and real-world benchmarks demonstrate that MFSR achieves efficient, flexible, and high-quality super-resolution, delivering results on par with or even better than multi-step teachers while requiring much lower computational cost.

Problem

• Real-World Image Super-Resolution
• One-Step Distillation
• Diffusion Models
• Flow-Based Models
• Inference Efficiency

Innovation

• MeanFlow Distillation
• One-step Super-Resolution
• Probability Flow ODE
• Classifier-Free Guidance
• Real-world Image Super-Resolution