Stochastic Deep Restoration Priors for Imaging Inverse Problems

TL;DR

ShaRP introduces a novel ensemble-based prior for inverse imaging problems, outperforming traditional denoising priors by effectively handling artifacts and enabling self-supervised training, with theoretical convergence guarantees.

Contribution

The paper presents ShaRP, a new method leveraging multiple restoration models as priors, improving inverse problem solutions over Gaussian denoisers and enabling self-supervised learning.

Findings

Achieves state-of-the-art results in MRI reconstruction.

Outperforms denoiser- and diffusion-based methods.

Supports self-supervised training without fully sampled data.

Abstract

Deep neural networks trained as image denoisers are widely used as priors for solving imaging inverse problems. While Gaussian denoising is thought sufficient for learning image priors, we show that priors from deep models pre-trained as more general restoration operators can perform better. We introduce Stochastic deep Restoration Priors (ShaRP), a novel method that leverages an ensemble of such restoration models to regularize inverse problems. ShaRP improves upon methods using Gaussian denoiser priors by better handling structured artifacts and enabling self-supervised training even without fully sampled data. We prove ShaRP minimizes an objective function involving a regularizer derived from the score functions of minimum mean square error (MMSE) restoration operators, and theoretically analyze its convergence. Empirically, ShaRP achieves state-of-the-art performance on tasks such as magnetic resonance imaging reconstruction and single-image super-resolution, surpassing both denoiser-and diffusion-model-based methods without requiring retraining.