DriftSR: One-Step Real-World Image Super-Resolution via Distribution Drifting

2026-10-03

AI Summary

This study addresses the optimization challenges inherent in existing single-step real-world image super-resolution methods, which typically rely on complex distillation or adversarial training. To this end, this work proposes DriftSR, a novel framework that leverages a frozen pre-trained diffusion prior via distribution drift to optimize solely a single-step generator. Specifically, it introduces spatial feature drift and structure modulation guidance mechanisms, eliminating the need for auxiliary encoders, distillation branches, or discriminators and thereby substantially streamlining the training pipeline. Extensive evaluations across three real-world benchmarks demonstrate that DriftSR achieves high-fidelity reconstruction while preserving the computational efficiency of single-step inference. Ultimately, this research provides a concise yet effective solution for one-step super-resolution.

Abstract

One-step real-world image super-resolution (Real-ISR) offers efficient inference, but recovering realistic and perceptually rich details often relies on score distillation or adversarial learning, introducing additional trainable components and making optimization more cumbersome. To this end, we propose DriftSR, a one-step Real-ISR framework that leverages pretrained diffusion priors through distribution drifting. Specifically, we perform drifting in the frozen intermediate representation space of a pretrained diffusion model, without introducing an additional task-specific feature encoder. Building on this space, we introduce Spatial Feature Drifting, which treats spatial features rather than entire images as distributional samples, enabling denser supervision for distribution alignment. To mitigate structural deviations, we further introduce Structure-Modulated Guidance, which adaptively refines drifting guidance according to local structural consistency with the LQ input. Consequently, DriftSR optimizes only the one-step generator, without auxiliary distillation branches or adversarial discriminators. Extensive experiments on three real-world benchmarks demonstrate that DriftSR delivers high-quality super-resolution reconstruction with efficient one-step inference.

Problem

• One-step image super-resolution
• Real-world image super-resolution
• Distribution alignment
• Optimization complexity

Innovation

• One-Step Super-Resolution
• Distribution Drifting
• Diffusion Prior
• Spatial Feature Drifting
• Structure-Modulated Guidance