Hyperspectral Imagery Band Selection Based on Maximal Standard Deviation

A new Hyperspectral image band selection algorithm based on maximal standard deviation is proposed to reduce spectral redundancy of Hyperspectral remote sensing image and computational complexity. It first uses standard deviation to measure the band information. The correlation between band and selecting band is then used as a weight factor for standard deviation in the iterative calculation. The experimental results show that the informative band subsets with low correlation can be selected using the maximal standard deviation. In addation, when the obtained bands are combined with Hyperspectral image classification, better classification accuracy can be produced.