Tensor Decomposition Based Multisensor Hyperspectral Fusion for Spatial Super Resolution

In this paper, we address the multi-sharpening problem by simultaneously fusing multiple multispectral images with a single hyperspectral image. We propose two variants of a novel strategy named G-STEREO-1 and G-STEREO-2. Both methods extend the tensor-based multi-sharpening framework named STEREO, by leveraging the complementarity of several multispectral sources to enhance fusion quality. G-STEREO-1 and G-STEREO-2 are based on a joint tensor decomposition model that incorporates a generalized Sylvester equation within a Canonical Polyadic (CP) tensor decomposition scheme. Our approaches overcome the limitations of existing joint tensor-based fusion techniques, which are restricted to fusing only a single multispectral image with a hyperspectral one. Experimental results show that both G-STEREO-1 and G-STEREO-2 consistently outperform these existing methods.