Discovering shape categories by clustering shock trees

Abstract

This paper investigates whether meaningful shape categories can be identified in an unsupervised way by clustering shock-trees. We commence by computing weighted and unweighted edit distances between shock-trees extracted from the Hamilton-Jacobi skeleton of 2D binary shapes. Next we use an EM-like algorithm to locate pairwise clusters in the pattern of edit-distances. We show that when the tree edit distance is weighted using the geometry of the skeleton, then the clustering method returns meaningful shape categories.