Perturbed conjugate gradient method: controlling semi-convergence, enforcing non-negativity, and accelerating iterations in tomographic imaging

Abstract

We regulate the semi-convergence behaviour, impose non-negativity on the iterations, reduce total variation, and enhance the conjugate gradient (CG) method by introducing a perturbed version of the CG method. We provide a convergence analysis for this perturbed version. We also demonstrate its performance using examples taken from tomographic imaging and compare it with CG and its projected version, superiorized conjugate gradient (S-CG-CD), and non-negative flexible CGLS (NN-FCGLS).