Extracting structure-tokens of cephalograms

Presents an algorithm and representation scheme for extracting structure tokens of cephalograms directly from X-ray images. The method is based on a principle of visual perception, i.e., edge partitioning and grouping which simulates human visual process on edge perception in certain aspects. With the method, image edges are first traced and partitioned into perceptual segments which are generic edge tokens. The edge tokens can be coded in a unique and descriptive format for both linear and curve segments. A grouping process is then performed to provide the joint tokens by gluing adjacent edge segments.