Texture Based Image Clustering Using Wavelets

Clustering is traditionally viewed as an unsupervised method for data analysis. The primary objective of cluster analysis is to partition a given data set into homogeneous clusters. In this paper, we present a novel algorithm for performing texture based clustering using wavelets. The approximation band of image Discrete Wavelet Transform is considered for segmentation which contains significant information of the input image. The Histogram based algorithm is used to obtain the number of regions and the initial parameters like mean, variance and mixing factor. The centroides are calculated and perform the clustering using k means. It is observed that the proposed method is computationally efficient than the k means algorithm and improved k means algorithm.