A comparison of regularization models for few-view CT image reconstruction

Abstract

In this paper I analyse some regularization models for the reconstruction of X-rays Computed Tomography images from few-view projections. It is well known that the widely used low-cost Filtered Back Projection method is not suitable in case of low-dose data, since it produces images with noise and artifacts. Iterative reconstruction methods based on the model discretization are preferred in this case. However, since the problem has infinite possible solutions and is ill-posed, regularization is necessary to obtain a good solution. Different iterative regularization methods have been proposed in literature, but an organized comparison among them is not available. We compare some regularization approaches in the case of few-view tomography by means of simulated projections from both a phantom and a real image.