GPU implementation of hyperspectral image classification based on weighted Markov random fields

Abstract

The dimensionality of hyperspectral data is very high, and spectral-spatial hyperspectral classification techniques are quite demanding from a computational viewpoint. In this paper, we present a computationally efficient implementation of a spectral-spatial classification method based on weighted Markov random fields. The method learns the spectral information from a sparse multinomial logistic regression (SMLR) classifier, and the spatial information is characterized by modeling the potential function associated with a weighted Markov random field (MRF) as a spatially adaptive vector total variation function. The parallel implementation has been carried out using commodity graphics processing units (GPUs) and the NVIDIA's compute unified device architecture (CUDA), thus exploiting the massively parallel nature of GPUs to achieve significant acceleration factors with regards to the serial version of the same classifier on an NVIDIA Tesla C2075 platform.