An iterative least squares approach to decorrelate minerals and ices contributions in hyperspectral images: Application to Cuprite (earth) and Mars

We present an iterative linear spectral unmixing model (ILSUM) which is aimed at finding the main surface components that contribute to the signal in visible and infrared hyperspectral images. We processed the global dataset of the OMEGA imaging spectrometer onboard Mars Express up to orbit 5300, covering two martian years. We also present a preliminary test on AVIRIS data on the Cuprite (Nevada) site. We use ILSUM to identify the contribution of each endmember of an input library containing laboratory spectra of ices and mineral powders that are representative of the main mineral families. Synthetic spectra (pure slope endmembers) are included to account at first order for aerosol and grain size variations. Applied to the global OMEGA data set, this algorithm provides a distribution map for the main minerals present on the martian surface, which appears to be mainly dominated by pyroxenes, olivine, ferric oxides, with localized exposures of sulfates and phyllosilicates.