Empirical mode decomposition based approach with spectral enhancement for hyperspectral image classification

Hyperspectral imaging systems provide high spectral information content by acquiring data in hundreds of narrow spectral bands. This high content of information results in a significant increase in classification accuracies for hyperspectral images with respect to optic or multispectral images. In this study, classification with empirical mode decomposition (EMD) and support vector machines (SVM), which has established its success in previous studies, is improved with spectral gradient enhancement for hyperspectral image classification. Intrinsic mode functions (IMFs) obtained by applying EMD to hyperspectral bands are combined using weights obtained by spectral gradient enhancement, resulting in high classification accuracies.