Based On Edge Extraction of ASM Automatic Landmark Placement

In the Active Shape Model, the most time consuming and scientifically unsatisfactory part of building shape models is the labeling of the training images. Manually placing hundreds (in 2D) of points on every image is both tedious and error prone. To reduce the burden, the combination of the image edge information and the traditional manual calibration methods have been developed. This method improves the calibration accuracy, and obtains more accurate statistical shape model and local texture model. Aiming at the characteristics of ASM modeling, this paper adopts a multiscale wavelet transform modulus maximum method of edge extraction, using the maximum variance method to obtain a threshold, after the use of connectivity judgment for each scale edge fusion. The simulation results show that, this algorithm can effectively reduce the burden, improve the modeling accuracy.