A Medical Image Registration Method Based on Gravity Optimization

The medical image registration method based on mutual information (MI) has the advantage of high precision, and not have to do any pretreatment. However, there may be local matches, or error brings by interpolation calculation, so the objective function contains many local maximums. In this paper, a novel image registration method, called Gravity Optimization (GO), is proposed for rigid registration of brain medical images, and compares the performance with Powell algorithm and Particle Swarm Optimization (PSO). The experiments of the 2D and 3D rigid registration of clinical brain images show that gravity optimization algorithm has a better performance at the rate of registration and global optimization.