Local binary patterns and Fourier transform based hyperspectral image classification

This paper proposes a novel classification framework based on a parallel architecture with Local Binary Pattern techniques and Fourier transform models. The Local Binary Pattern model is used as a descriptor employed on the spatial domain of the hyperspectral data. The Fourier transform method is applied along the spectral dimension of the hyperspectral set for each pixel in order to evaluate the data in terms of magnitude spectrum. Our proposed model is based on the concatenation of these types of spectral and spatial descriptors, for a more accurate classification of the data. The experimental results demonstrate that the proposed method can improve the accuracy on Indian Pines and Pavia University sets even with a limited number of training samples with a reduced computation time.