Hubness Implementation for High Dimensional Data Clustering Using Image Feature Extraction

Many data domains such as personal data, image data have large number of attributes which is difficult to cluster using traditional data mining techniques due to high dimensionality of data. Image clustering is one of the essential problems in automatic image processing, since high dimensional data exhibit high hubness which is data points appears at high density area of data. We find that image data set under several feature can be represented and cluster using hubness. We propose a novel approach of hubness in clustering of high dimensional data like images. Each feature of image including its resolution will treat as dimension of the image and using these all dimensions we apply clustering using hub concept. Hub is the data point that frequently occurs in k- nearest neighbor of other data points. This hub point can be used effectively as centroid in cluster prototype which will considerably speed up the convergence of algorithm.