A K-means shape classification algorithm using shock graph-based edit distance

Abstract

Skeleton is a very important feature for shape-based image classification. In this paper, we apply the discrete shock graph-based skeleton features to classify shapes into predefined groups, using a k-means clustering algorithm. The graph edit cost obtained by transforming database image graph into the respected query graph, will be used as distance function for the k-means clustering. To verify the performance of the suggested algorithm, we tested it on MPEG-7 dataset and our algorithm shows excellent performance for shape classification. © 2010 Springer-Verlag Berlin Heidelberg.