An Improved K2DPCA Dimensional Reduction method for Hyper spectral Remote Sensing Image
Feng Hui1,2, Pan Zijin1
COMPUTER MODELLING & NEW TECHNOLOGIES 2013 17(5B) 96-99
1Faculty of Mechanical Electronic and Information, Jiangsu Polytechnic of finance &Economics, No.8, Meicheng East Road, Huai’an, China
2College of Computer and Information, Hohai University, No.1, Xikang Road, Nanjing, China
An improved kernel two-dimensional principle analysis (K2DPCA) dimensional reduction method for hyperspertral remote sensing image was proposed in this paper. It decorrelated the columns of remote sensing image by the standard K2DPCA, then used columns 2DPCA to further decorrelate the row direction. It could achieve the dimensional reduction at both widthways and lengthways for remote sensing image. The original images could be reconstructed by the principle components of extracted from each bands of remote sensing image. Experiments were verified with AVIRIS hyperspertral remote sensing image Cuprite, and the result showed that this new method could not only ensure the reconstructed image quality, but also effectively improve the image compression rate.