A joint band prioritization and band-decorrelation approach to band selection for hyperspectral image classification

Band selection for remotely sensed image data is an effective means to miti gate the curse of dimensionality, Many criteria have been suggested in the past for optimal band selection, In this paper, a joint band-prioritization and band-decorrelation approach to band selection is considered for hypers pectral image classification. The proposed band prioritization is a method based on the eigen (spectral) decomposition of a matrix from which a loadin g-factors matrix can be constructed for band prioritization via the corresp onding eigenvalues and eigenvectors, Two approaches are presented, principa l components analysis (PCA)-based criteria and classification-based criteri a. The former includes the maximum-variance PCA and maximum SNR PCA, wherea s the latter derives the minimum misclassification canonical analysis (MMCA ) (i.e., Fisher's discriminant analysis) and subspace projection-based crit eria. Since the band prioritization does not take spectral correlation into account, an information-theoretic criterion called divergence is used for band decorrelation, Finally, the band selection can then be done by an eige nanalysis-based band prioritization in conjunction with a divergence-based band decorrelation, It is shown that the proposed band-selection method eff ectively eliminates a great number of insignificant bands. Surprisingly, th e experiments show that with a proper band selection, less than 0.1 of the total number of bands can achieve comparable performance using the number o f full bands, This further demonstrates that the band selection can signifi cantly reduce data volume so as to achieve data compression.