SURFACE REGISTRATION FOR THE ANALYSIS OF 3D ORTHODONTIC TREATMENT

Abstract

Image registration is a crucial step in image analysis tasks in which the final information is gained from the combination of various data sources like in image fusion, change detection, and multi-channel image restoration. Typically, registration is required in medicine, to obtain more complete information about the patient, such as monitoring tumor growth, treatment verification, and comparison of the patient's data with anatomical atlases. Registration techniques can be based on a limited set of identified points, on the alignment of segmented structures, most commonly object surfaces, or directly onto measures computed from the image intensity values. This work is concerned with the development of a new method for the surface registration of three-dimensional (3D) objects with a special interest in skull registration. Two main approaches are introduced to find the optimal registration parameters. The first approach is based on the iterative closest point (JCP) algorithm and the second is based on matching surface signature images. In both approaches, at first, some points are chosen and known as critical points. The critical points are selected depending on three different criteria. The criteria are based on the distance histogram, the neighborhood condition and the principal curvatures. The different criteria resulted in various critical points and accordingly different registration accuracies. The results have shown that the one with the highest performance is the method based on the principal curvatures calculation. In the signature images approach, surface signature images are created at each critical

point. By matching these images from both objects the registration parameters are found. Various image matching techniques are presented. In the iterative approach, the JCP algorithm is applied only on the critical points chosen instead of applying it to the whole point set. This significantly has reduced the processing time while preserving almost the same accuracy of applying the ICP algorithm alone.