A Background Correction Algorithm for Hyperspectral Images

This paper introduces a method for background (or baseline) correction in hyperspectral images. The method is based on the optimization of a criterion incorporating a non-quadratic robust loss (data fidelity term) and both spatial and spectral regularization terms to enforce baseline smoothness. Unlike the classical approach based on a pixel-by-pixel baseline correction, the proposed algorithm exploits jointly the spatial and spectral information. The effectiveness of the the proposed algorithm is demonstrated using simulated and geological hyperspectral images.