Spectrally Coarse-To-Fine Pansharpening For Hyperspectral Images

Hyperspectral (HS) pansharpening synthesizes low spatial resolution HS (LRHS) images and connected panchromatic (PAN) images to form high spatial resolution HS (HRHS) images. A major problem that such a task is faced with is the huge spectral gap between the two kinds of data. To handle that, in this work we propose a spectrally coarse-to-fine network (SCFNet) to recover the HRHS images, which combines the LRHS images and the PAN images to build multiple low spectral resolution HRHS images in several spectral resolution levels, and then fuse them into the final HRHS images. Experiments are conducted to compare the proposed method with some of representative techniques, where the superiority of our method is validated.