Comparison of sparse-view CT image reconstruction algorithms

Abstract

In recent years, the restoration of computerized tomography (CT)images with low-dose projection is a key issue in CT image processing. The sparse views-based methods have been proposed to achieve reasonable image quality. This paper studies three conventional sparse-view CT image reconstruction algorithms: the total variational minimization projection onto convex set (TVM-POCS) algorithm, the two-step iterative Shrinkage-Thresholding (TwIST) algorithm, and the iterative filtered back projection (FBP) algorithm. The three algorithms are compared and analyzed in terms of the computational complexity, universal quality index(UQI), and structure similarity index(SSIM). Two experiments with comparison are performed in the case of sparse-view and low-dose projection, respectively. The computed results reveal that under Poisson noise environments, the TVM-POCS algorithm has superior performance over other algorithms in restoration quality and computing time.