Anatomically-Aware, Automatic, and Fast Registration of 3D Ear Impression Models
Alexander Zouhar, Tong Fang, Gozde Unal, Hui Xie, Greg Slabaugh and Fred McBagonluri
We present a registration framework based on feature points of anatomical 3D shapes represented in the point cloud domain. Anatomical information is utilized throughout the complete registration process. The surfaces, which in this paper are ear impression models, are considered to be similar in the way that they possess the same anatomical regions but with varying geometry. First, in a shape analysis step, features of important anatomical regions (such as canal, aperture, and concha) are extracted automatically. Next these features are used in ordinary differential equations that update rigid registration parameters between two sets of feature points. For refinement of the results, the GCP algorithm is applied. Through our experiments, we demonstrate our technique’s success in surface registration through registration of key anatomical regions of human ear impressions. Furthermore, we show that the proposed method achieves higher accuracy and faster performance than the standard GCP registration algorithm.
Third International Symposium on 3D Data Processing, Visualization and Transmission (3DPVT) 2006