Anisotropic total variation for limited-angle CT reconstruction

Abstract

The CT limited-angle problem has been discussed for many years because of the strong demand in medical imaging applications. However as a severely ill-posed problem, it is hard to achieve accurate reconstruction if no additional prior information is given. Recently, an ART+TV reconstruction method was proposed. With the prior knowledge of image sparsity, the ART+TV gives high quality reconstructions far beyond other methods. In our work, we further explored the ART+TV method on limited-angle problems and delivered a new ART+ATV method. In our new method, the actual range of view angles is considered as another prior information and according to it an anisotropic total variation (ATV) norm is designed to perform L1-minimization. Meanwhile inherited from the ART+TV method, the prior knowledge of image sparsity is considered as well. With the combination of the two priors, the ART+ATV method shows a high-quality and fast-convergence reconstruction on limited-angle problems.