Fast Hyperspectral Subspace identification using Eigenvalue based energy thresholding

Blind hyperspectral unmixing comprises three subproblems. First is subspace identification that is finding the number of pure endmembers present. Second endmember extraction and the third is abundance estimation. There are many approaches for endmember extraction and abundance estimation in literature. But out of that many approaches requires prior knowledge of the number of endmembers. We are proposing a new approach EET (Eigenvalue based Energy Thresholding) for hyperspectral subspace identification. This proposed approach is best suitable for real-time application as computation time is low compared to other approaches. We have compared our algorithm results with other state-of-art algorithms on real and synthetic datasets.