MSDRN: Pansharpening of Multispectral Images via Multi-Scale Deep Residual Network

In order to acquire a high resolution multispectral (HRMS) image with the same spectral as (MS) and spatial panchromatic (PAN) image, pansharpening, typical hot fusion topic, has been well researched. Various pansharpening methods that are based on convolutional neural networks (CNN) different architectures have introduced by prior works. However, scale information of source images is not considered these methods, which may lead loss high-frequency details in fused image. This paper proposes method MS via multi-scale deep residual network (MSDRN). The proposed constructs multi-level make better use images. Moreover, learning into further improve ability feature extraction simplify process. A series experiments conducted QuickBird GeoEye-1 datasets. Experimental results demonstrate MSDRN achieves superior or competitive performance state-of-the-art both visual evaluation quantitative evaluation.