A Spectral Projected Gradient optimization for binary tomography

Abstract

In this paper we present a deterministic binary tomography reconstruction method based on the Spectral Projected Gradient (SPG) optimization approach. We consider a reconstruction problem with added smoothness convex prior. Using a convex-concave regularization we reformulate this problem to a non-integer and box constrained optimization problem which is suitable to solve by SPG method. The flexibility of the proposed method allows application of other reconstruction priors too. Performance of the proposed method is evaluated by experiments on the limited set of artificial data and also by comparing the obtained results with the ones provided by the often used non-deterministic Simulated Annealing method. The comparison shows its competence regarding to the quality of reconstructions. © 2010 Springer-Verlag Berlin Heidelberg.