A Stability Benchmark of Generative Regularizers for Inverse Problems

TL;DR

This paper evaluates the stability and reliability of generative priors, especially diffusion models, in inverse imaging problems, comparing them with optimization-based methods across various stability criteria.

Contribution

It introduces a comprehensive numerical benchmark for stability properties of generative priors and compares their performance with traditional optimization techniques.

Findings

Generative priors perform well in certain stable settings.

They may struggle with out-of-distribution data and model inaccuracies.

Benchmark results highlight scenarios where generative methods excel or fall short.

Abstract

Generative (diffusion) priors demonstrate remarkable performance in addressing inverse problems in imaging. Yet, for scientific and medical imaging, it is crucial that reconstruction techniques remain stable and reliable under imperfect settings. Typical definitions of stability encompass the notion of ''convergent regularization'', robustness to out-of-distribution data, and to inaccuracies in the forward operator or noise model. We evaluate these properties numerically. Furthermore, we benchmark generative approaches against modern optimization-based methods inspired by the widely used variational techniques. Our results give insights for which settings and applications generative priors can deliver state-of-the-art reconstructions, and on those in which they fall short or may even be problematic.