Estimation of number of endmembers in a Hyperspectral image using Eigen thresholding

Unmixing of Hyperspectral images is essentially a blind or semi-blind source separation problem that tries to extract object (endmember) spectra and abundance from the image which is considered as a linear or non-linear mixture of spectrally distinct objects (endmembers) weighted by fractional abundance. An accurate and efficient method for estimation of number of endmembers present in the image scene is the starting point in spectral unmixing. Various unmixing and endmember determination methods have assumed that the number of endmembers is known from prior knowledge about the image scene. This paper presents an Eigen analysis based method to determine number of endmember present in an image by separating signal subspace and noise subspace. Signal components and noise components in a linear noisy mixture are characterized by the dominant Eigen values of covariance matrix and the low Eigen values of covariance matrix respectively. In this paper number of endmembers is estimated by separation of signal and noise components by thresholding. Results obtained in synthetic and real images shows that this scheme gives accurate estimation in low noise levels.