Extended K-means with an Efficient Estimation of the Number of Clusters