The application of wavelet theory to cephalogrm landmarking

The purpose of this paper is to apply the wavelet transform to an automated landmarking system for extracting the landmarks in the PA view cephalogram. The cephalograms in P.A views obtained from National Cheng Kung University Hospital are then digitized using a PC-based image acquisition system for later analysis. With the image processing techniques and the predefined facial boundary, a simple edge detector is designed for detecting the reference point and the width of face. Afterwards, the image is then divided into the rectangular subimages containing the landmarks. In this study, the first derivative of a Gaussian function is used as the wavelet function to extract the edges from the multi-scales of the images. Based upon the definitions of anatomy, the concept of least distance, Hough transform, and the reflex auxiliary line from the landmarks at the cephalogram in LAT view, the position of landmarks are located in PA view cephalogram so as to reduce the errors and time resulting from the manual measurements. From the results, it is shown that the average errors for the landmarks extracted using this proposed system are less than 4mm.