Multiband image fusion using total generalized variation regularization

Multiband image fusion has become a thriving area of research in a number of different fields, such as space robotics, and remote sensing, etc. Many multiband image fusion methods have been proposed for hyperspectral sharpening with panchromatic images, hyperspectral sharpening with multispectral images and panchromatic images, etc. Despite the different motivations, we observe that many existing methods possibly lead to over-smooth regions. In this work, we consider a new problem formulation of two image fusion problems. A novel fusion model with total generalized variation regularization term is proposed, where the fusion process is performed on hybrid gradient domains. The optimization framework of alternating direction multiplier of method is used to solve the resulting problem. In an extensive evaluation, our method outperforms some state-of-the-art methods.