Abstract
Pan-sharpening is a significant task in remote sensing image processing, which merges a high-resolution panchromatic (PAN) image and a low-resolution multispectral (MS) image to create a high-resolution MS image. In this article, we propose a novel deep-learning-based MS image pan-sharpening method that combines a shallow–deep convolutional network (SDCN) and a spectral discrimination-based detail injection (SDDI) model. SDCN consists of a shallow network and a deep network, which can capture mid-level and high-level spatial features from PAN images. SDDI, inspired by the “Amelioration de la Resolution Spatial par Injection de Structures” concept, is developed to merge the spatial details extracted by SDCN into MS images with minimal spectral distortion. SDCN and SDDI are collaboratively learned for achieving high-spatial-resolution MS image and preserving more spectral information. Both the visual assessment and the quantitative assessment results on IKONOS and QuickBird datasets confirmed that the proposed method outperforms several state-of-the-art pan-sharpening methods.
Highlights
P ANCHROMATIC (PAN) images and multispectral (MS) images are two types of remote sensing images acquired simultaneously by an optical satellite
These compared algorithms are from different categories of pan-sharpening methods: GS is based on component substitution; Indusion is one of the multiresolution analysis methods; SR is based on sparse representation learning technology; pansharpening neural network (PNN) is based on a convolutional neural network (CNN) containing three layers; PanNet is a residual network; and MSDCNN is a multiscale feature extraction method based on a CNN
We proposed a novel deep-learning-based MS image pan-sharpening approach via combining the advantages of shallow–deep network learning and spectral discrimination-based detail injection (SDDI) model
Summary
P ANCHROMATIC (PAN) images and multispectral (MS) images are two types of remote sensing images acquired simultaneously by an optical satellite. A residual network (ResNet) for pan-sharpening was proposed in PanNet [29], and a two-branch network with a deeper structure was proposed in RSIFNN [30] that can separately capture the salient features of the MS and PAN images Most of those studies emphasize feature extraction without considering the difference of spatial details in each band. SDDI is proposed to inject the extracted spatial details into each band of the upsampled MS images In this way, the SDDI preserves spectral characteristics and improves fusion performance simultaneously.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
