Multimodal image registration based on weighted feature and mutual information

In this paper, a weighted feature based image registration scheme has been proposed. The registration is carried out in the feature space of the CT and MR images. In the registration process different weights have been assigned to different modalities. The registration of the images have been achieved by maximizing the weighted normalized mutual information of the feature CT and MR images. For a given image, the feature image is generated by weighting the features extracted from each image. It has been observed that the variation of MI over the feature space is smoother than that of using original image.