A priorconditioned LSQR algorithm for linear ill-posed problems with edge-preserving regularization

TL;DR

This paper introduces a priorconditioned LSQR algorithm for efficiently solving large-scale linear inverse problems with edge-preserving regularization, accelerating convergence using priorconditioning embedded into the forward operator.

Contribution

It develops a matrix-free, factorization-free priorconditioned LSQR method that incorporates prior information directly into the forward operator for improved convergence.

Findings

Effective in 3D fluorescence diffuse optical tomography

Accelerates convergence of Krylov methods with priorconditioning

Demonstrates efficiency with algebraic multigrid preconditioner

Abstract

This article presents a method for solving large-scale linear inverse problems regular- ized with a nonlinear, edge-preserving penalty term such as the total variation or Perona-Malik. In the proposed scheme, the nonlinearity is handled with lagged diffusivity fixed point iteration which involves solving a large-scale linear least squares problem in each iteration. Because the convergence of Krylov methods for problems with discontinuities is notoriously slow, we propose to accelerate it by means of priorconditioning. Priorconditioning is a technique which embeds the information contained in the prior (expressed as a regularizer in Bayesian framework) directly into the forward operator and hence into the solution space. We derive a factorization-free priorconditioned LSQR algorithm, allowing implicit application of the preconditioner through efficient schemes such as multigrid. The resulting method is matrix-free i.e. the forward map can be defined through its action on a vector. We demonstrate the effectiveness of the proposed scheme on a three-dimensional problem in fluorescence diffuse optical tomography using algebraic multigrid preconditioner.