Hyperspectral And Multispectral Image Fusion Based On Deep Attention Network

Hyperspectral (HS) images have rich spectral information and can provide attribute information. High spatial resolution images, such as multispectral (MS) images and panchromatic (PAN) images, can provide fine geometric features. Thus, the fusion of the two images can achieve information complementarity and increase the accuracy and reliability of information. In this paper, we propose a hyperspectral and multispectral image fusion method based on deep attention network. Our model consists of two parts. One is the fusion network, which is used to fuse images. The other part is the spatial attention network, which is used to extract tiny textures and enhance the spatial structure. Experimental results compared with some state-of-the-art methods illustrate that our method is outstanding in both visual and numerical results.