Medial Axis Extraction Using Growing Neural Gas

In this paper, a medial axis extracting method is proposed by using a self-organising neural network, namely the Growing Neural Gas (GNG), which is applied to preserve the topology of any input space. The boundaries of a given figure are sampled with induced Delaunay triangulation corresponding to the set of nodes in GNG network at the first stage; then the Voronoi diagram for these neurons are constructed and the approximation of the medial axes are extracted from the Voronoi diagram; finally, the global medial axes is obtained by pruning the useless branches. Simulations are carried out to demonstrate the performance of the proposed method. Experimental results show that GNG method is able to extract global medial axis effectively and has the resistance ability to small deviation of the boundary of the figure.