Plug-and-Play Methods for Integrating Physical and Learned Models in Computational Imaging

TL;DR

This paper reviews Plug-and-Play Priors (PnP), a framework that combines physical sensor models and learned regularizers for advanced computational imaging reconstructions across various applications.

Contribution

It provides a unified, principled review of PnP, tracing its origins, variations, main results, and applications, and discusses recent equilibrium-based developments.

Findings

PnP achieves state-of-the-art imaging reconstructions.

Applications include bio-microscopy, CT, MRI, and ptycho-tomography.

Recent work links PnP to equilibrium equations.

Abstract

Plug-and-Play Priors (PnP) is one of the most widely-used frameworks for solving computational imaging problems through the integration of physical models and learned models. PnP leverages high-fidelity physical sensor models and powerful machine learning methods for prior modeling of data to provide state-of-the-art reconstruction algorithms. PnP algorithms alternate between minimizing a data-fidelity term to promote data consistency and imposing a learned regularizer in the form of an image denoiser. Recent highly-successful applications of PnP algorithms include bio-microscopy, computerized tomography, magnetic resonance imaging, and joint ptycho-tomography. This article presents a unified and principled review of PnP by tracing its roots, describing its major variations, summarizing main results, and discussing applications in computational imaging. We also point the way towards further developments by discussing recent results on equilibrium equations that formulate the problem associated with PnP algorithms.