Gradient Domain Diffusion Models for Image Synthesis
This addresses efficiency issues in generative image synthesis for applications in image processing, computer vision, and machine learning, though it appears incremental.
The paper tackles the slow convergence of diffusion models in image synthesis by performing diffusion in the gradient domain, which is mathematically equivalent but sparser, resulting in faster convergence as confirmed by numerical experiments.
Diffusion models are getting popular in generative image and video synthesis. However, due to the diffusion process, they require a large number of steps to converge. To tackle this issue, in this paper, we propose to perform the diffusion process in the gradient domain, where the convergence becomes faster. There are two reasons. First, thanks to the Poisson equation, the gradient domain is mathematically equivalent to the original image domain. Therefore, each diffusion step in the image domain has a unique corresponding gradient domain representation. Second, the gradient domain is much sparser than the image domain. As a result, gradient domain diffusion models converge faster. Several numerical experiments confirm that the gradient domain diffusion models are more efficient than the original diffusion models. The proposed method can be applied in a wide range of applications such as image processing, computer vision and machine learning tasks.