A new improved method of mutual information combined with edge correlation deviation

This article studies the principle of the maximum mutual information method in gray-scale multi-modal medical image registration. On the basis of mutual information, the goal function of registration is improved. This paper integrates mutual information with edge correlation deviation and defines a new registration measure function. Image registration is finished through optimizing this function. This method not only makes use of gray-scale mutual information of original images but also contains the pixel relationship of edge images in two-dimensional level. It plays roll of compensation for drawback of mutual information. Experiments prove that this is an effective image registration method. © 2009 Copyright SPIE - The International Society for Optical Engineering.