Learning Image Priors through Patch-based Diffusion Models for Solving Inverse Problems
This addresses the bottleneck of high computational cost and data requirements for diffusion models in inverse problems, particularly for high-resolution domains like medical imaging, though it is incremental as it builds on existing diffusion inverse solvers.
The paper tackles the computational and data inefficiency of training diffusion models for high-dimensional inverse problems by proposing PaDIS, a patch-based method that learns image priors from patches, achieving improved memory and data efficiency while outperforming previous methods with limited training data in tasks like CT reconstruction and deblurring.
Diffusion models can learn strong image priors from underlying data distribution and use them to solve inverse problems, but the training process is computationally expensive and requires lots of data. Such bottlenecks prevent most existing works from being feasible for high-dimensional and high-resolution data such as 3D images. This paper proposes a method to learn an efficient data prior for the entire image by training diffusion models only on patches of images. Specifically, we propose a patch-based position-aware diffusion inverse solver, called PaDIS, where we obtain the score function of the whole image through scores of patches and their positional encoding and utilize this as the prior for solving inverse problems. First of all, we show that this diffusion model achieves an improved memory efficiency and data efficiency while still maintaining the capability to generate entire images via positional encoding. Additionally, the proposed PaDIS model is highly flexible and can be plugged in with different diffusion inverse solvers (DIS). We demonstrate that the proposed PaDIS approach enables solving various inverse problems in both natural and medical image domains, including CT reconstruction, deblurring, and superresolution, given only patch-based priors. Notably, PaDIS outperforms previous DIS methods trained on entire image priors in the case of limited training data, demonstrating the data efficiency of our proposed approach by learning patch-based prior.