Mutual Information of High Area Scale Space Images

Medical image registration methods based on the maximization of mutual information have shown promising results. But it needs much computing time. This paper presents a new method to rigid registration based on the combination of scale images and mutual information. Firstly, the concept of area morphology is introduced. And then, the scale images are constructed with the area morphological operators. Not only the noise in the image is removed, but also the intensity levels are decreased in the scale image by area morphological operators. Moreover, the images do not be blurred. As a result, the matching process is accelerated while the accuracy and the robustness of the method is hold. For magnetic resonance and computed tomography images, behavior of mutual information matching is examined in a scale scheme. The normalized mutual information and the area scale space are combined in registration for the first time. Experimental results show that the scale approach to mutual information matching works well.