Parallel computation of mutual information in multicore environment & its applications in medical image registration

Abstract

Mutual information (MI) is the measure of amount of information containing by one random variable about another. It is an established similarity measure used for image registration. Calculation of mutual information is very computationally meticulous and time consuming at the time of computation of the immense images or at the time of processing with set of images. Main purpose of this research work is to propose a proficient method to compute mutual information used for image registration using parallel computing. This method is able to work with different numbers of threads to take all the benefits of the processors having multiple cores like core i3, core i5, core i7 after maintaining the synchronization between cores. Our experiment shows the significant speed up using easily available multi core architectures especially for core i7 processor. Another major purpose of this paper is to describe the application of the parallel computation of mutual information in image registration especially in the case of medical image registration like CT (Computed Tomography) - PET (Positron Emission Tomography) and Other modalities of Neuroimaging. © 2014 IEEE.