UltraImage: Rethinking Resolution Extrapolation in Image Diffusion Transformers

TL;DR

UltraImage introduces a frequency correction and adaptive attention method to significantly improve high-resolution image generation in diffusion transformers, enabling extrapolation beyond training scales with reduced repetition and enhanced quality.

Contribution

The paper presents a novel frequency correction technique and entropy-guided attention mechanism to address resolution extrapolation challenges in image diffusion transformers.

Findings

Outperforms prior methods on Qwen-Image and Flux datasets.

Enables generation of images up to 6K*6K resolution.

Reduces content repetition and improves visual fidelity.

Abstract

Recent image diffusion transformers achieve high-fidelity generation, but struggle to generate images beyond these scales, suffering from content repetition and quality degradation. In this work, we present UltraImage, a principled framework that addresses both issues. Through frequency-wise analysis of positional embeddings, we identify that repetition arises from the periodicity of the dominant frequency, whose period aligns with the training resolution. We introduce a recursive dominant frequency correction to constrain it within a single period after extrapolation. Furthermore, we find that quality degradation stems from diluted attention and thus propose entropy-guided adaptive attention concentration, which assigns higher focus factors to sharpen local attention for fine detail and lower ones to global attention patterns to preserve structural consistency. Experiments show that UltraImage consistently outperforms prior methods on Qwen-Image and Flux (around 4K) across three generation scenarios, reducing repetition and improving visual fidelity. Moreover, UltraImage can generate images up to 6K*6K without low-resolution guidance from a training resolution of 1328p, demonstrating its extreme extrapolation capability. Project page is available at \href{https://thu-ml.github.io/ultraimage.github.io/}{https://thu-ml.github.io/ultraimage.github.io/}.