Dimensional Reduction of Hyperspectral Image Data Using Band Clustering and Selection Based on Statistical Characteristics of Band Images

In this paper an approach for the dimensionality reduction of the hyperspectral image data using the method of band selection based on the statistical measures is introduced. The spread hyperspectral image data is measured in each band and the calculated bands are clustered using the K-means clustering technique. The K-means clustering of bands is performed in such a way that the intra-cluster variance is kept minimize and the inter-cluster variance maximum. The optimal number of band selection is done using the concept of Virtual Dimensionality (VD). The endmember or targets are extracted through Vertex Component Analysis (VCA). The experimental results are compared with other unsupervised band selection techniques to show the effectiveness of the proposed technique.