Medical Image Registration Using Normal Vector and Intensity Value

Medical image registration plays a vital role in many applications such as diagnosis and image guided intervention. So far, the mainstream of medical image registration algorithms are based on intensity values of the images, eg. mutual information. However, spatial feature of the image is another important component deserved to be considered. In this paper, we propose a novel metric combing spatial and intensity scale information, which lies on the similarity features of normal vector and intensity value of the corresponding localization of two aligned images. The experiments reveal that the proposed metric achieved better accuracy comparing with the metric based only on intensity information.