Investigation on accelerated ordered subsets image reconstruction techniques with superiorization methodology

Abstract

In this paper, we consider a perturbed (superiorized) version of ordered subsets simultaneous iterative reconstruction technique (OS-SIRT) aiming to find a superior solution to linear system of equations which is usually ill-conditioned and has noise in its right-hand side. Furthermore, to speed up the iterations, we apply a recently introduced generalized relaxation scheme and give a convergence analysis based on the algorithmic operators. The performance and computational gain of the proposed algorithm are illustrated by data generated by simulating a computed tomography problem.