Extended study of k-Means clustering technique for human face classification and recognition

In this paper we have presented an extended study of k-Means clustering technique for human face classification and recognition as well. To execute the same a classification technique is presented based on k-means algorithm. Calculating the cluster values for each image matrix turned vector; then compiling a composite matrix for a complete training data methods are proposed based on summative, quantitative approaches and calculating difference vectors. Throughout the paper justification has been cited on behalf of using k-Means algorithm for clustering. It also has been seen during the extended study on k-Means that decent recognition rates can be achieved as experiments have been done on ORL database and Facial Expression Database - Japanese Female (JAFFE) with our proposed approaches and achieved recognition rate of 90% and 85% respectively with less computation time.