Graph Regularized Nonnegative Matrix Factorizationfor Hyperspectral Data Unmixing

Spectral unmixing is an important tool inhyperspectral data analysis for estimating endmembers andabundance fractions in a mixed pixel. This paper examines theapplicability of a recently developed algorithm called graphregularized nonnegative matrix factorization (GNMF) for thisaim. The proposed approach exploits the intrinsic geometricalstructure of the data besides considering positivity and fulladditivity constraints. Simulated data based on the measuredspectral signatures, is used for evaluating the proposedalgorithm. Results in terms of abundance angle distance (AAD)and spectral angle distance (SAD) show that this method caneffectively unmix hyperspectral data.

Keywords:

Hyperspectral Imagery , Linear Mixing Model(LMM) , Spectral Unmixing , Graph Regularized NonnegativeMatrix Factorization (GNMF)