PSNR Based Optimization Applied To Maximum A Posteriori Expectation Maximization For Image Reconstruction On A Multi-Core System

Abstract

Statistical reconstruction methods present high potential image quality as compared to analytical methods; however, it suffers of time complexity. To reduce reconstruction time statistical reconstruction algorithm such as Maximum a Posteriori via Expectation Maximization algorithm (MAPEM) is parallelized in a shared memory processing (SMP) environment. This work exposes a parallel MAPEM algorithm that reconstructs an image on a multi-core parallel environment to reduce the execution time. An attempt to optimize the iteration required to reconstruct an image in various angle is performed. The execution time and speed up and efficiency factors for both serial and parallel MAPEM are computed. The present work uses phantom data sets of various sizes under different number projections. The research exhibits that the parallel computing environment provides the source of high computational power leading to reconstruct an image instantaneously.