One-Step Diffusion Transformer for Controllable Real-World Image Super-Resolution

2025-11-21

AI Summary

In real-world image super-resolution (Real-ISR), balancing fidelity and controllability remains challenging: multi-step diffusion methods offer high diversity but suffer from low fidelity, whereas single-step approaches achieve strong fidelity yet lack flexible control. To address this, we propose ODTSR, a single-step diffusion Transformer model. Its core innovation is the Noise-Mixed Visual Stream (NVS) architecture, which employs dual-path inputs—encoding a degradation-controllable image with adjustable noise and a prior-preserving image with fixed noise—and integrates Fidelity-Aware Adversarial (FAA) training to jointly optimize reconstruction accuracy and perceptual consistency. ODTSR is the first method enabling controllable super-resolution for complex content—including Chinese text and natural images—without scene-specific fine-tuning. On general Real-ISR benchmarks, it achieves state-of-the-art performance, significantly improving fine-detail recovery and visual naturalness.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Mixed Discrete/Continuous Search
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Data and user privacy-enhancing technologies for the Web
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Recent advances in diffusion-based real-world image super-resolution (Real-ISR) have demonstrated remarkable perceptual quality, yet the balance between fidelity and controllability remains a problem: multi-step diffusion-based methods suffer from generative diversity and randomness, resulting in low fidelity, while one-step methods lose control flexibility due to fidelity-specific finetuning. In this paper, we present ODTSR, a one-step diffusion transformer based on Qwen-Image that performs Real-ISR considering fidelity and controllability simultaneously: a newly introduced visual stream receives low-quality images (LQ) with adjustable noise (Control Noise), and the original visual stream receives LQs with consistent noise (Prior Noise), forming the Noise-hybrid Visual Stream (NVS) design. ODTSR further employs Fidelity-aware Adversarial Training (FAA) to enhance controllability and achieve one-step inference. Extensive experiments demonstrate that ODTSR not only achieves state-of-the-art (SOTA) performance on generic Real-ISR, but also enables prompt controllability on challenging scenarios such as real-world scene text image super-resolution (STISR) of Chinese characters without training on specific datasets.

Problem

• Balancing fidelity and controllability in image super-resolution
• Overcoming generative randomness in multi-step diffusion methods
• Enabling prompt controllability without dataset-specific training

Innovation

• One-step diffusion transformer for super-resolution
• Noise-hybrid visual stream with adjustable noise
• Fidelity-aware adversarial training for enhanced controllability