Unsupervised feature extraction for hyperspectral images using combined low rank representation and locally linear embedding

Abstract

Hyperspectral images(HSIs) provide hundreds of narrow spectral bands for the land-covers, thus can provide more powerful discriminative information for the land-cover classification. However, HSIs suffer from the curse of high dimensionality, therefore dimension reduction and feature extraction are essential for the application of HSIs. In this paper, we propose an unsupervised feature extraction method for HSIs using combined low rank representation and locally linear embedding (LRR LLE). The proposed method can simultaneously use both the spectral and spatial correlation within HSIs, with LRR modelling the intrinsic property of union of low-rank subspaces and LLE considering the correlation within spatial neighbours. Experiments are conducted on real HSI datasets and the classification results demonstrate that the features extracted by LRR LLE are more discriminative than the state-of-art methods.