Detecting Critical Points of Skeletons Using Triangle Decomposition of Gradient Vector Flow Field

Skeletons of an object have a variety of applications, such as automatic path generation, shape matching, identification, and classification. In this paper, we introduce a method to detect the critical points of skeletons, then skeletons of an object can be generated by linking the detected critical points. The method is based on concepts from 2-D phase plane analysis of the linear Gradient Vector Flow (GVF) field, which is generated from the original image. The critical points on the skeleton are automatically found by searching each triangle generated from decomposition of the whole field. The critical points with small magnitudes in the valley areas of the GVF are located by an affine-invariant detector. Some preliminary results are shown to demonstrate the performance of the proposed method.