HYPERSPECTRAL IMAGE FUSION

ABSTRACT: In this research, we jointly process high spectral and high geometric resolution images, exploiting their synergies, with aims to (a) generate a fused image of high spectral and geometric resolution, and (b) to improve spectral unmixing level of hyperspectral images at subpixel regarding the estimation of endmembers and fractional abundances. The potential of this methodological improvement lies in a more accurate and detailed semantic inter-pretation of objects and their properties in hyperspectral and multispectral images, with applications often in environ-mental mapping, monitoring and change detection. To incorporate the case of real data the spectral response of the one sensor with respect to the other must be known. To estimate the relative spectral response the two images must be co-registered and the point spread function must be taken into account. In this paper, we degrade the images spectrally and geometrically so that they have the same dimensions and we solve a least-squares problem with two constraints. The results using synthetic data confirm this approach, while with real data results are in principle promising, but in some cases the reconstructed spectral response is noisy. 1