BEMD and wavelet denoising based classification for hyperspectral image

摘要

A high-accuracy algorithm based on combination of bi-dimensional empirical mode decomposition (BEMD) and wavelet denoising is presented in this paper, in which BEMD is adapted to decompose optimal bands selected from feature selection technique into many bi-dimensional intrinsic mode functions (BIMFs) and sym4 wavelet is chosen to denoise these BIMFs, so that the denoised BIMFs could be taken as input of support vector machine (SVM). Experimental results indicate that the proposed approach not only has promising accuracy but also significantly reduces complexity and computational time of SVM.