Endmember extraction algorithms for hyperspectral image based on independent component analysis

This paper presents a new algorithm for endmember extraction on hyperspectral images based on independent component analysis. ICA is a recent technique used to tackle the blind source separation problem, which mixed signals need to be separated without knowing the mixing matrix and the source signals. Based on the assumption of the distribution of endmembers being independent, we transfer the problem of endmember extraction to the BSS problem, and a joint diagonalization algorithm is used to solve the BSS problem. The effectiveness of the algorithm has been verified by the simulation.