High accuracy hyperspectral image classification based on Empirical Mode Decomposition and composite kernel

This paper proposes to use empirical mode decomposition (EMD) to increase the classification accuracy of hyperspectral images. EMD is a nonlinear and adaptive signal decomposition approach and decomposes signals into intrinsic mode functions (IMFs) and a final residue. In this paper, initially, EMD is applied to each hyperspectral image band and the IMFs corresponding to each hyperspectral image band are obtained. Then, the information contained in the first IMFs and second IMFs of each band are combined using composite kernels. Support vector machine (SVM) based classification is used to show the classification performance of the proposed approach. Experimental results show that the SVM classification accuracy can significantly be improved using the proposed EMD and composite kernel based classification approach.