FISTA Iterates Converge Linearly for Denoiser-Driven Regularization

Abstract

The effectiveness of denoising-driven regularization for image reconstruction has been widely recognized. Two prominent algorithms in this area are Plug-and-Play ($\text{PnP}$) and Regularization-by-Denoising ($\text{RED}$). We consider two specific algorithms $\text{PnP-FISTA}$ and $\text{RED-APG}$, where regularization is performed by replacing the proximal operator in the $\text{FISTA}$ algorithm with a powerful denoiser. The iterate convergence of $\text{FISTA}$ is known to be challenging with no universal guarantees. Yet, we show that for linear inverse problems and a class of linear denoisers, global linear convergence of the iterates of $\text{PnP-FISTA}$ and $\text{RED-APG}$ can be established through simple spectral analysis.