COMPARISON OF CONVENTIONAL AND BISECTING K-MEANS ALGORITHMS ON COLOR QUANTIZATION

摘要

Color quantization represents an important technique withboth many applications, as well as outstanding challengesin the field of image processing. Numerous clustering techniqueshave been applied to the task of achieving the optimalquantization with varying results. In this paper, wecompare the results of the conventional k-means algorithm,indisputably the most celebrated nonhierarchical clusteringalgorithm, to those of a recent hierarchical variant ofit named bisecting k-means. We show that, while bisectingk-means often produces slightly higher quantization distortionthan k-means, the former more than makes up for thisin stability and computational efficiency.