A Wavelet Diffusion GAN for Image Super-Resolution

2024-10-23

arXiv.org

AI Summary

To address the slow training and inference of diffusion models in single-image super-resolution (SISR), hindering real-time deployment, this paper proposes a wavelet-domain conditional Diffusion GAN framework. It innovatively integrates discrete wavelet transform (DWT) into the diffusion process, enabling conditional generation and multi-scale residual learning directly in the wavelet domain—thereby significantly reducing feature dimensionality and backward diffusion steps. By synergistically combining the high-fidelity synthesis capability of diffusion models with the efficient modeling strength of GANs, our method achieves new state-of-the-art performance on CelebA-HQ: it improves PSNR and SSIM substantially, accelerates training by 2.3×, and reduces per-image inference time to 18 ms for ×4 SISR. To our knowledge, this is the first approach to enable high-fidelity, real-time super-resolution.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

In recent years, diffusion models have emerged as a superior alternative to generative adversarial networks (GANs) for high-fidelity image generation, with wide applications in text-to-image generation, image-to-image translation, and super-resolution. However, their real-time feasibility is hindered by slow training and inference speeds. This study addresses this challenge by proposing a wavelet-based conditional Diffusion GAN scheme for Single-Image Super-Resolution (SISR). Our approach utilizes the diffusion GAN paradigm to reduce the timesteps required by the reverse diffusion process and the Discrete Wavelet Transform (DWT) to achieve dimensionality reduction, decreasing training and inference times significantly. The results of an experimental validation on the CelebA-HQ dataset confirm the effectiveness of our proposed scheme. Our approach outperforms other state-of-the-art methodologies successfully ensuring high-fidelity output while overcoming inherent drawbacks associated with diffusion models in time-sensitive applications.

Problem

• Improving real-time feasibility of diffusion models for super-resolution
• Reducing training and inference time with wavelet-based GAN
• Enhancing high-fidelity image output in time-sensitive applications

Innovation

• Wavelet-based conditional Diffusion GAN
• Reduces reverse diffusion timesteps
• Uses DWT for dimensionality reduction