Analysis of Clustering Algorithms for Image Segmentation and Numerical Databases

Abstract

Clustering techniques are broadly used in research are as where pattern recognition is needed, like in signal processing, automatic voice analysis, computer vision, and data mining. However, for each specific problem, the adequate technique must be selected in order to achieve better results. In this paper, a comparative analysis between the three mostly used clustering techniques (k-means, ISODATA, and the sequential clustering algorithm) is presented. The goal of the analysis is to compare the efficiency of each algorithm applied to numerical databases and images. The results of the application of the algorithms to a set of 25 images (natural and artificial) and 5 numerical databases are presented and discussed.