Provably Convergent Plug-and-Play Quasi-Newton Methods

Hong Ye Tan, Subhadip Mukherjee, Junqi Tang, Carola-Bibiane Schönlieb

Research output: Contribution to journalArticlepeer-review

53 Downloads (Pure)

Abstract

Plug-and-Play (PnP) methods are a class of efficient iterative methods that aim to combine data fidelity terms and deep denoisers using classical optimization algorithms, such as ISTA or ADMM, with applications in inverse problems and imaging. Provable PnP methods are a subclass of PnP methods with convergence guarantees, such as fixed point convergence or convergence to critical points of some energy function. Many existing provable PnP methods impose heavy restrictions on the denoiser or fidelity function, such as non-expansiveness or strict convexity, respectively. In this work, we propose a novel algorithmic approach incorporating quasi-Newton steps into a provable PnP framework based on proximal denoisers, resulting in greatly accelerated convergence while retaining light assumptions on the denoiser. By characterizing the denoiser as the proximal operator of a weakly convex function, we show that the fixed points of the proposed quasi-Newton PnP algorithm are critical points of a weakly convex function. Numerical experiments on image deblurring and super-resolution demonstrate 2--8x faster convergence as compared to other provable PnP methods with similar reconstruction quality.
Original languageEnglish
Pages (from-to)785-819
JournalSIAM Journal on Imaging Sciences
Volume17
Issue number2
DOIs
Publication statusPublished - 2 Apr 2024

Fingerprint

Dive into the research topics of 'Provably Convergent Plug-and-Play Quasi-Newton Methods'. Together they form a unique fingerprint.

Cite this