Diffusion Model for Image Generation - A Survey
Xinrong Hu, Yuxin Jin, Jinxing Liang, Junping Liu, Ruiqi Luo, Min Li, Tao Peng · 2023
Diffusion models are a class of generative model that excels in generating high-quality images, making them the state-of-the-art among other generative models. Their impressive image generation capabilities and diverse generation patterns have garnered significant attention. To facilitate the widespread utilization of diffusion models in various domains, this article provides a comprehensive overview of their applications in the field of image generation. Firstly, we introduce the foundational theory of diffusion models and the widely adopted general frameworks. Then, we categorize the applications of diffusion models in the field of image generation into the following areas: unconditional generation, conditional generation, text-to-image translation, image-to-image translation, image super-resolution, image editing, and image inpainting techniques. Furthermore, we thoroughly discuss the limitations of diffusion models in the context of image generation and provide insights into future research and development directions in this field.