Speed Up the Image Registration with PDE Model and Parallel AOS Scheme

Image registration is to find a suitable space transformation that aligns two or more images of the same scene taken at different times. As the image size is increasing, the numerical approaches based on the pixel matrix may stepup the computation complexity, and it is difficult to perform the large size registration on a single computer in the limited time. PDE-based registration is employing the partial differential equation driven by a similarity measure to minimize the distance between the corresponding pixel points. During the minimization process, Additive operator splitting (AOS) scheme is usually be used. In this paper, we discuss a multithreading method to perform AOS scheme and apply it to the parallel image registration computing with OpenMP and MPI on a dual core cluster. Some experimental results show that this method can produce the large size parallel image registration and save the time consume efficiently.