An Efficient Band Selection Method For Hyperspectral Imageries Based On Covariance Matrix

Band selection plays an important role in reducing the dimensionality of hyperspectral data sets. It is a combinatorial optimization problem for optimal band (feature) subset selection which generally involves high computational complexity. In this paper, we present an efficient band selection methods based on the covariance matrix. The method tries to compute the subset of bands with the largest determinant of covariance matrix. By using a recursive role found in this study, the proposed method can be very efficient. Besides, it also has the potentiality to determine the number of required bands.