Ranking-Based Band Selection Using Correlation and Variance Measure

Contiguous narrow bands of hyperspectral images greatly increase computational complexity. Redundancy reduction is therefore necessary. In this chapter, a minimum redundancy– and maximum variance–based unsupervised band selection method is presented. Since ranking-based band selection methods are iterative in nature, the huge spatial dimension of the hyperspectral image increases the computation time of the dimensionality reduction (DR) method. Hence, in this ranking-based band selection method, discrete wavelet transformation (DWT) is applied on the data to reduce the spatial redundancy without much affecting the overall band correlations. This in turn made the process more time efficient and noise resilient. Highly correlated bands are considered similar, and the one with higher variance is accepted as being more discriminating. Finally, the selected bands are classified, and overall accuracy (OA) is calculated. This method is compared with four other existing state-of-the-art methods in the similar field in terms of OA and execution time for evaluating the performance.