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

Ruiqing Wang, Yuanzhi Zhu, Kai Zhang, Hanshu Yan, Shilin Lu, Jian Yang

Keywords: Diffusion models

TL;DR: We use MeanFlow Distillation to achieve one step real-world image super-resolution.

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, high-fidelity results in a single step while still allowing an optional multi-step path for further improvement. Our approach uses MeanFlow as the learning target, enabling the student to approximate the mean 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 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 better than multi-step teachers while requiring much lower computational cost.

Primary Area: generative models

Submission Number: 534