A Multi-stage Approach for 3D Teeth Segmentation from Dentition Surfaces

In this paper, we present a multi-stage approach for teeth segmentation from 3D dentition surfaces based on a 2D model-based contour retrieval algorithm. First, a 3D dentition model is loaded to the system and a range image is created. Second, binarized 2D sectional images are generated and contours are extracted. During several processing steps a set of tooth contour candidates are produced and they are evaluated. The best-fitting contour for each tooth is refined using snakes. Finally, the 2D contours are integrated to full 3D segmentation results. Due to its excellent experimental results, our algorithm has been applied in the practical realization of a so-called virtual articulator currently being developed for dentistry. Today, only mechanical articulators are applied in the dental practice. They are used in the fabrication and testing of removable prosthodontic appliances (dentures), fixed prosthodontic restorations (crowns, bridges, inlays and onlays), and orthodontic appliances. Virtual articulators are supposed to simulate the same functionality, however, in a much more flexible and convenient way.