Hyperspectral and multispectral image fusion using local spatial-spectral dictionary pair

This paper deals with the spatial resolution enhancement of low resolution (LR) hyperspectral image (HSI) by fusing it with high resolution (HR) multispectral image (MSI) of the same observed scene. A new HSI and MSI fusion approach based on local spatial-spectral dictionary pair is proposed. In the proposed approach, HR MSI and its spatial degradation version (LR MSI) are divided into overlapped subimages for the purpose of HR and LR dictionary pair construction. To incorporate spatial and spectral information simultaneously, spatial-spectral dictionary is generated rather than the traditional spectral or spatial ones. Meanwhile, a localized strategy is employed for dictionary construction to generate a local dictionary rather than a global one, to reduce the computational cost. By appropriately assuming that the LR HSI and desired HR HSI can be collaboratively represented by LR dictionary and HR dictionary respectively sharing the same set of representation coefficients, the desired HR HSI is reconstructed by HR dictionary and the obtained collaborative representation coefficients. Simulative experimental results illustrate that the proposed HSI and MSI fusion approach is capable of producing better or comparable fused results compared with some state-of-the-art fusion approaches using spectral or spatial dictionaries, with much lower computational cost. This makes it quite promising in practical application.