UltraPixel: Advancing Ultra-High-Resolution Image Synthesis to New Peaks

TL;DR

UltraPixel introduces a cascade diffusion model architecture that efficiently generates ultra-high-resolution images (1K to 6K) with detailed quality, reduced complexity, and shared parameters, advancing the state-of-the-art in high-res image synthesis.

Contribution

The paper presents UltraPixel, a novel multi-resolution diffusion model with implicit neural representations and scale-aware normalization, enabling efficient, high-quality ultra-high-resolution image synthesis within a single model.

Findings

Achieves state-of-the-art performance in high-resolution image synthesis.

Reduces training complexity and data requirements.

Maintains high image quality across multiple resolutions.

Abstract

Ultra-high-resolution image generation poses great challenges, such as increased semantic planning complexity and detail synthesis difficulties, alongside substantial training resource demands. We present UltraPixel, a novel architecture utilizing cascade diffusion models to generate high-quality images at multiple resolutions (\textit{e.g.}, 1K to 6K) within a single model, while maintaining computational efficiency. UltraPixel leverages semantics-rich representations of lower-resolution images in the later denoising stage to guide the whole generation of highly detailed high-resolution images, significantly reducing complexity. Furthermore, we introduce implicit neural representations for continuous upsampling and scale-aware normalization layers adaptable to various resolutions. Notably, both low- and high-resolution processes are performed in the most compact space, sharing the majority of parameters with less than 3$\%$ additional parameters for high-resolution outputs, largely enhancing training and inference efficiency. Our model achieves fast training with reduced data requirements, producing photo-realistic high-resolution images and demonstrating state-of-the-art performance in extensive experiments.