Non-negative matrix factorization pansharpening of hyperspectral data: An application to mid-infrared astronomy

Mid-infrared (wavelengths of 2-25μm) astronomy has progressed significantly in the last decades, thanks to space and ground based telescopes. Space observatories benefit from the absence of atmospheric absorption, allowing to reach the very high sensitivities needed to perform 3D hyperspectral observations at relatively low angular resolution (4”). On the other hand, ground based facilities that suffer from strong atmospheric absorption make use of large (above 8m diameter) telescopes to perform sub-arcsecond resolution imaging through selected windows in the mid-infrared range. In this Paper, we present a method based on non-negative matrix factorization to merge data from space and ground based mid-IR telescopes in order combine the best sensitivity, spectral coverage and angular resolution. We prove the efficiency of this technique on real mid-IR astronomical data, and suggest that it can be applied to any hyper-spectral astronomical data-set.