Parallel Implementation for AOS Scheme on a Dual-core Cluster

As the image size increases, it is difficult to perform the large size image processing on a single computer in a limited time. One of the most efficient solving methods is parallel computing. Most numerical approaches for image computing with nonlinear diffusion filtering by means of an unconditional stable semi-implicit schemes are implemented based on an additive operator splitting (AOS) scheme. The basic idea of AOS scheme is split the whole process in a sequence of 1-D processes, so it can be performed by a group of threads. In this paper, we discuss a multithreading method to perform AOS scheme and apply it to the parallel image registration computing with OpenMP on a dual core cluster. First, we discuss the diffusion equation and its AOS scheme, then detail three parallel implementation with multithreads program ming using OpenMP and MPI. Some experimental results show that this method can produce the large size parallel image registration and save the time consume efficiently.