Spectral library pruning method in hyperspectral sparse unmixing

Abstract

Sparse unmixing algorithm aims at finding the optimal subset of signatures from a spectral library to best model each pixel in hyperspectral image and estimating their corresponding abundance. However, the high mutual coherence of spectral library limits the performance of sparse unmixing algorithm. In this paper, a method referencing the extracted information from hyperspectral image was managed to prune spectral library (REiPSL) for improving the accuracy and efficiency of sparse unmixing. The REiPSL method was applied to the simulated hyperspectral data and Airborne Visible Infrared Imaging Spectrometer (AVIRIS) image. The experiments show that the accuracy and efficiency are improved by comparing with the results obtained using the complete spectral library directly.