CT Image Reconstruction using Fourth-order PDE based AD Regularization

Abstract

One-Step-Late (OSL) statistical iterative algorithm plays an important role in Computed Tomography (CT) image reconstruction. But the fundamental problem related to OSL is the optimum initial value condition, slow convergence, and ill-posed. To resolves these problems, we present a modified OSL algorithm. The issue of optimum initial value condition and slow convergence can handle by integrating the Simultaneous Algebraic Reconstruction Technique (SART) with OSL called as modified OSL (SART+OSL). The output of modified OSL undertakes in Fourth order partial differential equation (PDE) based Anisotropic Diffusion regularization approach to deal with an ill-posed. It is an extended version of the Perona-Malik (P-M) filter. For validation of the proposed model, both simulated and real standard thorax phantoms have been used. Finally, the results were compared with the related state-of-the-art methods. It is observed that the proposed model has many desirable advantages such as noise reduction, minimize the computational cost, as well as accelerate the convergence rate.