Pattern Recognition, 4th Edition

15.4. Binary Morphology Clustering Algorithms (BMCAs)

Algorithms of this type are suitable for cases in which clusters are not properly represented by a single representative ([Post 93, Mora 00]). The idea here is to map X to a discrete set S that facilitates the clustering procedure and then use the identified clusters in S as a guide for the identification of the clusters in X. In the sequel, we describe such an algorithm, called the binary morphology clustering algorithm [Post 93]. The BMCA involves four main stages. During the first stage the data set is discretized and a new set is derived. This is the so-called discrete binary (DB) set. During the second stage, the basic morphological operators (opening and closing) are applied on the ...