PCA Based Regional Mutual Information for Robust Medical Image Registration

摘要

Mutual information (MI) has emerged in recent years as an effective similarity metric measuring registration of images. One drawback of MI is that it just takes into account only the relationships between corresponding pixels, that is, the spatial information of images is not utilized. To overcome this problem, regional mutual information (RMI) has been proposed and demonstrated that it effectively includes the spatial information of images and much robust than MI. However, RMI has to spend too much computing time for calculating a high dimension joint distribution. For improvement, we propose a novel PCA based regional mutual information (PRMI) to implement a more robust and faster medical image registration.