The J-SparseFI-HM Hyperspectral Resolution Enhancement Method - Now Fully Automated.

Abstract

This paper introduces the new scheme of the previously proposed Joinly Sparse Fusion of Hyperspectral and Multispectral Imagery (J-SparseFI-HM) fusion method. This extended, now fully automated and parallelized version of J-SparseFI-HM jointly estimates bundles of an adjustable number of high resolution hyperspectral bands by fusing corresponding low resolution bands with possibly multiple high resolution multispectral ones. Which multispectral bands are individually used is decided via a decision matrix that is calculated from the spectral response functions of the multispectral sensor. Tests are performed on the SuperMUC petascale system. Recently acquired 0.75 m airborne VNIR HySpex data is used to synthesize a WorldView-2 image as well as low resolution hyperspectral data with a down-sampling factor of 10. The fusion results are compared to those produced by three state-of-the-are hyperspectral resolution enhancement methods.