Endmember finding and spectral unmixing using least-angleregression

摘要

A new endmember finder and spectral unmixing algorithm based on the LARS/Lasso method for linear regres-sion is developed. The endmember finder is sequential; a single endmember is identified at first and furtherendmembers which depend on the previous ones are found. The process terminates once a pre-determined num-ber of endmembers have been found, or when the modeling error has attained the noise floor. The unmixingalgorithm is a straightforward procedure that expresses each pixel as a linear combination of endmembers in aphysically meaningful way. This algorithm successfully unmixes simulated data, and shows promising results onreal hyperspectral images as well.