Abstract
This paper proposes an image reconstruction algorithm which combines the Relativity-of-Gaussian (RoG) as a regularization with Simultaneous Algebraic Reconstruction Technique algorithm (SART). RoG model can distinguish noise from edges/structures and identify potential edges, thereby it can effectively eliminate noise without distorting edges/structures. It can minimize the consequences from ringing-effect on the edge caused by the Gibbs phenomenon of the local filters, and deal with the problems affected by noise at which the global optimization methods (TV, WLS, RTV) fail. With simulation experiments on the Shepp-Logan head model and a simulation data, this paper gives the comparison results of sparse-angle reconstruction. In both cases, it shows that the reconstructed images of the SART-RoG algorithm are better than the others.