Modified Vector Quantization using Cluster Validity Measure

Clustering is an important approach in engineering and other scientific disciplines. One clustering approaches is known as Vector Quantization (VQ). The main disadvantage of vector quantization algorithm is that the number of codevectors must be supplied as a parameter. In this paper we propose a Modified Vector Quantization (MVQ) algorithm which it uses Davies-Bouldin validity measure. This measure allows the number of codevectors to be determined automatically. We applied the MVQ technique for image segmentation and results are satisfactory.

Keywords:

Clustering , LBG , Vector quantization , Cluster Validity , Davies-Bouldin index