Estimating the Intrinsic Dimension of Hyperspectral Images Using an Eigen-Gap Approach

摘要: 3AbstractLinear mixture models are commonly used to represent hyperspectral datacube as a linearcombinations of endmember spectra. However, determining the number endmembers for imagesembedded in noise is crucial task. This paper proposes fully automatic approach estimating thenumber images. The estimation based on recent results randommatrix theory related so-called spiked population model. More precisely, we study gapbetween successive eigenvalues sample covariance matrix constructed from high dimensionalnoisy samples. resulting strategy unsupervised and robust correlated noise. Thisstrategy validated both synthetic real experimental very promisingand show accuracy this algorithm with respect state-of-the-art algorithms.Index TermsHyperspectral imaging, linear spectral mixture, number, random theory,sample matrix, eigen-gap approach.