BINARY IMAGE REGISTRATION BASED ON GEOMETRIC MOMENTS: APPLICATION TO THE REGISTRAION OF 3D SEGMENTED CT HEAD IMAGES

In this paper, we present a binary image registration strategy for the registration of segmented CT scan data of the human head. For the characterization of the 3D skull binary images we adopt a powerful representation of the binary images: the geometric moment invariants (GMIs). They provide pertinent and discriminant information related to the geometry of the binary objects, thereby leading to match points which have similar geometric properties, and also to enhance the quality of the matching process. For the registration algorithm we propose to use the topology-preserving B-spline-based registration method proposed by Noblet et al. for the registration of MRI head images. The algorithm has proven its efficiency and high registration precision. Since the information carried out by the different GMIs is complementary, we study and discuss the importance of combining various GMIs to better guide the matching process. Results obtained using synthetic deformation fields highlight the promising performance of the strategy.