文摘
Extraction of craniofacial anatomical structures on cephalometric radiographs is important for observing and predicting the growth changes and evaluating the curative effect in orthodontic analysis. Due to the natural of cephalograms, it is difficult for simplex image processing approaches to exactly track both the soft tissue contours and the structure outlines of internal skull. In this paper, we address this problem by recognizing 262 landmark points and then connecting them according to the prior knowledge. Image processing, pattern matching, and modified active shape model are combined to realize the landmark recognition; subdivision curves are used to obtain the structure rendering.