Ultra-High-Resolution Image Synthesis: Data, Method and Evaluation

2025-06-02

AI Summary

Ultra-high-definition (4K) image synthesis has long been hindered by the absence of standardized benchmarks and computational bottlenecks. To address this, we introduce Aesthetic-4K—the first standardized dataset specifically designed for 4K synthesis—and propose Diffusion-4K, an end-to-end high-fidelity generative framework. Our key contributions include: (1) a novel Scale-Consistent VAE jointly optimized with Wavelet-based Latent Fine-tuning (WLF) to balance reconstruction fidelity and inference efficiency; (2) two new evaluation metrics—GLCM Score and Compression Ratio—that jointly quantify texture detail preservation and global structural fidelity; and (3) integration of GPT-4o–assisted automatic annotation, multi-scale assessment (GLCM/CLIPScore/FID), and compatibility with large foundation models (e.g., Flux-12B). Experiments demonstrate a 32% reduction in FID and a 41% improvement in GLCM Score over state-of-the-art methods. Both code and dataset are publicly released.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Ultra-high-resolution image synthesis holds significant potential, yet remains an underexplored challenge due to the absence of standardized benchmarks and computational constraints. In this paper, we establish Aesthetic-4K, a meticulously curated dataset containing dedicated training and evaluation subsets specifically designed for comprehensive research on ultra-high-resolution image synthesis. This dataset consists of high-quality 4K images accompanied by descriptive captions generated by GPT-4o. Furthermore, we propose Diffusion-4K, an innovative framework for the direct generation of ultra-high-resolution images. Our approach incorporates the Scale Consistent Variational Auto-Encoder (SC-VAE) and Wavelet-based Latent Fine-tuning (WLF), which are designed for efficient visual token compression and the capture of intricate details in ultra-high-resolution images, thereby facilitating direct training with photorealistic 4K data. This method is applicable to various latent diffusion models and demonstrates its efficacy in synthesizing highly detailed 4K images. Additionally, we propose novel metrics, namely the GLCM Score and Compression Ratio, to assess the texture richness and fine details in local patches, in conjunction with holistic measures such as FID, Aesthetics, and CLIPScore, enabling a thorough and multifaceted evaluation of ultra-high-resolution image synthesis. Consequently, Diffusion-4K achieves impressive performance in ultra-high-resolution image synthesis, particularly when powered by state-of-the-art large-scale diffusion models (eg, Flux-12B). The source code is publicly available at https://github.com/zhang0jhon/diffusion-4k.

Problem

• Lack of standardized benchmarks for ultra-high-resolution image synthesis
• Computational constraints in generating ultra-high-resolution images
• Need for comprehensive evaluation metrics for ultra-high-resolution images

Innovation

• Aesthetic-4K dataset with GPT-4o captions
• Diffusion-4K framework with SC-VAE and WLF
• GLCM Score and Compression Ratio metrics