HiWave: Training-Free High-Resolution Image Generation via Wavelet-Based Diffusion Sampling

2025-06-25

AI Summary

Diffusion models face significant challenges in ultra-high-resolution (UHR) image generation, including prohibitive computational cost and poor zero-shot cross-resolution generalization, often resulting in artifacts such as object duplication and structural distortion. To address these issues, we propose a fine-tuning-free, two-stage, training-free framework: first, block-wise DDIM inversion extracts low-frequency structural priors; second, high-frequency guidance is explicitly incorporated in the wavelet domain to enforce multi-scale detail consistency during sampling. The method operates entirely on off-the-shelf pre-trained diffusion models (e.g., SDXL) without introducing any additional parameters or retraining. Experiments demonstrate substantial artifact suppression and robust spatial coherence and texture fidelity at resolutions ≥1024×1024. A user study confirms strong perceptual preference, with over 80% of participants favoring our outputs—validating both superior generation quality and enhanced perceptual plausibility.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Generation

Application Category

• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Diffusion models have emerged as the leading approach for image synthesis, demonstrating exceptional photorealism and diversity. However, training diffusion models at high resolutions remains computationally prohibitive, and existing zero-shot generation techniques for synthesizing images beyond training resolutions often produce artifacts, including object duplication and spatial incoherence. In this paper, we introduce HiWave, a training-free, zero-shot approach that substantially enhances visual fidelity and structural coherence in ultra-high-resolution image synthesis using pretrained diffusion models. Our method employs a two-stage pipeline: generating a base image from the pretrained model followed by a patch-wise DDIM inversion step and a novel wavelet-based detail enhancer module. Specifically, we first utilize inversion methods to derive initial noise vectors that preserve global coherence from the base image. Subsequently, during sampling, our wavelet-domain detail enhancer retains low-frequency components from the base image to ensure structural consistency, while selectively guiding high-frequency components to enrich fine details and textures. Extensive evaluations using Stable Diffusion XL demonstrate that HiWave effectively mitigates common visual artifacts seen in prior methods, achieving superior perceptual quality. A user study confirmed HiWave's performance, where it was preferred over the state-of-the-art alternative in more than 80% of comparisons, highlighting its effectiveness for high-quality, ultra-high-resolution image synthesis without requiring retraining or architectural modifications.

Problem

• High-resolution image synthesis without training
• Reducing artifacts in zero-shot generation
• Enhancing structural coherence and fine details

Innovation

• Training-free wavelet-based diffusion sampling
• Two-stage pipeline with DDIM inversion
• Wavelet-domain detail enhancer for coherence