Comparative Analysis of Hyperspectral Dimension Reduction Methods

Hyperspectral sensors generate images in narrow bands in continuous manner with hundreds of spectral bands. The data with large number of bands require more processing power to classify. To decrease the redundancy in hyperspectral images and increase classifying efficiency with less number of bands, dimension reduction techniques are applied. In this paper, linear and non-linear dimension reduction methods are compared in classification performance and calculation time.