HyperSpectral classification with adaptively weighted L1-norm regularization and spatial postprocessing

Sparse regression methods have been proven effective in a wide range of signal processing problems such as image compression, speech coding, channel equalization, linear regression and classification. In this paper we develop a new method of hyperspectral image classification based on the sparse unmixing algorithm SUnSAL for which a pixel adaptive L1-norm regularization term is introduced. To further enhance class separability, the algorithm is kernelized using a RBF kernel and the final results are improved by a combination of spatial pre and post-processing operations. We show that our method is competitive with state of the art algorithms such as SVM-CK, KLR-CK, KSOMP and KSSP.