ABSTRACT
Spectral unmixing (SU) of hyperspectral images (HSIs) has been in the spotlight of both research and applications during the recent years [1]. In a nutshell, SU refers to the process of (a) detecting the pure material spectra (called endmembers) that are present in the scene and (b) estimating their corresponding fractional proportions in each pixel (called abundances). This unmixing process is commonly addressed in literature as two distinct tasks. That said, several endmembers’ extraction algorithms have been proposed in literature, (e.g. [2,3]) focusing on the identification of the materials’ spectra. On the contrary, abundance estimation algorithms [4,5], presuppose that the dictionary of the endmembers is already known, concentrating on the accurate estimation of the abundance values.
