Bilinear Matrix Factorization using a Gradient Method for Unmixing Hyperspectral Images Combinedc with Multispectral Data

In this investigation, a new method is proposed for bilinearly unmixing hyperspectral images combined with multispectral data. The designed algorithm, which is used to unmix the considered hyperspectral image, is based on bilinear matrix factorization with nonnegativity constraints. Using new gradient-based iterative update rules, it optimizes a cost function that includes multispectral data and a spectral degradation model between these data and hyperspectral ones. The initialization stage of the proposed algorithm also uses the considered multispectral variables. Simulation tests, based on realistic synthetic data, reveal performance superiorities of the proposed method compared to literature approaches.