CT reconstruction from few-views by edge guided TV minimization

Abstract

Based on compressed sensing (CS) recovery theory, total variation (TV) minimization has been successfully used in computed tomography (CT) reconstruction for sparse or limited angle data. When the number of projection views is much smaller or noise exists in the projection data, a conventional TV minimization algorithm often suffers from the decrease of spatial resolution especially in the edge area. Considering that the edge is an important index for image quality and it reflects the sparsity of an image to some extent, in this paper, we propose an edge guided TV (EGTV) minimization reconstruction algorithm for better edge preservation in CT reconstruction. EGTV with both isotropic and anisotropic weights of the TV discretization term is derived by importing edge information into TV calculating process. When an edge of the to-be-estimated image is detected, the associated weight of the TV additive element is adjusted. To solve the EGTV minimization reconstruction problem, a similar TV-based minimization implementation was developed to deal with the raw data fidelity and other constraints. The results with computer simulation reveals that EGTV minimization algorithm can improve the image quality and preserve the edge characteristics compared to conventional TV minimization algorithm.