Linear unmixing of hyperspectral images for analysis of fluorescently-labeled cellswith imperfect endmember spectra

Abstract

Spectral unmixing is the method of the detecting and localizing subpixel features by estimating the relative concentrations of the reference spectra. For most applications, spectral unmixing methods should account for spectral reference ambiguity, and concentration estimates with non-negativity and sum-to-one constraints. In this paper, we propose total least squares (TLS) based methods for unmixing of hyperspectral images obtained via fluorescence microscopy. Here, we formulate the restricted TLS as a constrained quadratic optimization problem which can be solved efficiently. The performance of restricted TLS is compared to the existing least squares based methods via simulations.