Pansharpening Algorithm For Multispectral/Hyperspectral Remote Sensing Images Based On The Intersecting Cortical Model

With the vigorous development of remote sensing technology,multi-source remote sensing images with different spatial and spectral resolutions are provided continuously by various earth observation platforms of satellites or aircrafts.Multi-source remote sensing images,contain as well as panchromatic images,multispectral images,hyperspectral images and drone remote sensing images,play a significant role in agricultural monitoring,environmental protection,military defense and other fields through their rich spatial and spectral information.However,due to limitations of the signal-to-noise ratio of remote sensors,it is difficult for a single satellite to obtain remote sensing images with both the high spatial resolution and the high spectral resolution.Therefore,it is necessary to develop spatial and spectral fusion algorithms to integrate complementary information from multi-source remote sensing images,which can make multi-source remote sensing images suitable for applications.Pansharpening is a technique,which injects spatial details of high-resolution images into low-resolution images to acquire a fusion remote sensing image with both high spatial and spectral resolutions.However,the classical pansharpening algorithms have limited improvement in spatial detail information and spectral fidelity.In particular,when pansharpening is applied to image fusion with large resolution differences of hyperspectral image,panchromatic image,and aerial image,there exists spectral mixing and the poor performance of spectral preservation and spatial detail injection.The intersecting cortical model is a laterally connected neural network of two-dimensional neurons that can obtain irregular segmentation regions consistent with human visual perception by its nonlinear activations of image pixels.Therefore,the paper aims at improving the pansharpening accuracy of multi-source remote sensing images by combining the ICM model with pansharpening methods of multispectral and hyperspectral remote sensing images.The main work and innovation points are as follows:(1)The paper proposes an ICM model based adaptive pansharpening algorithm for multispectral and panchromatic images.Most of the traditional pansharpening algorithms always use the entire image or regular regions to calculate the fusion injection coefficient,which will result in the mixing between the foreground and the background to affect the quality of the fused image.The paper proposes a novel adaptive segmentation pansharpening algorithm combined with the ICM model.The algorithm takes advantages of the nonlinear segmentation characteristics of the ICM which adaptively calculates the weight of detail injection.It also uses the shuffled frog leaping algorithm to optimize the parameters of the ICM.Finally,the sharpened multispectral image with spectral fidelity and rich spatial details is obtained.(2)The paper proposes an adaptive pansharpening algorithm for multispectral and UAV images based on the ICM model.In the paper,we inject the sub-meter spatial resolution information of the UAV images to the multispectral image through a step-by-step fusion method.At the same time,the ICM irregular segmentation regions,which conform to human visual characteristics,are used to calculate the injection weight in the fusion.The algorithm improves the spatial resolution of multispectral images to the sub-meter level of the UAV image,while maintaining the spectral information of the multispectral image.(3)The paper proposes an adaptive pansharpening algorithm for hyperspectral images based on the ICM model.Since the problems of large differences in spatial and spectral resolution will lead to the deficiency of spectral distortion and texture detail missing in hyperspectral pansharpening.The paper proposes an adaptive pansharpening algorithm for hyperspectral images based on the ICM model,which uses a step-by-step fusion strategy with the multispectral image as the medium.At the same time,the gray wolf optimizer algorithm is used to adaptively optimize the ICM parameters to generate optimal irregular segmented regions.The proposed algorithm can acquire more accurate details and spectral information for the hyperspectral images.The experiments are carried out by using WorldView-2,GF-2,GF-1 and ZY-1 02 D datasets.The experimental results show that the proposed algorithms are superior to the other algorithms both in the evaluation indices of the spatial details and the spectral information.The experimental results have verified the effectiveness of the proposed algorithm.