Research On Multi-source Remote Sensing Image Fusion Algorithm

Image fusion of multi-source remote sensing is to combine two or more remote sensing images of the same scene that obtained by different sensors,in order to gain a new image which is richer and more complete in information.Image fusion is not only the vital part of remote sensing detection,but also has been applied in many fields,such as environmental detection,precise agriculture,and urban planning.In recent years,with the development of satellite remote sensing technology,scholars have done extensive researches on image fusion algorithms of multi-source remote sensing.According to the image fusion level,image fusion can be divided into pixel-level fusion,feature-level fusion,and decision-level fusion.In this paper,the research is mainly focused on multispectral and panchromatic images which are based on the pixel-level fusion,aiming for suppressing the spectral distortion while enhancing space characteristic information of image.The specific work is as follows:(1)An image fusion algorithm based on adaptive fractional differential is proposedTo solve the problem that image edge details are not fully considered in image fusion based on IHS transform,an image fusion algorithm based on adaptive fractional differential is proposed.The basic idea is to use up-sampling operation into multispectral image firstly,then the multispectral image is transformed by IHS transform to obtain intensity component.Secondly,panchromatic image and intensity component are enhanced by adaptive fractional differential,which can improve the edge details and keep the flat region.Next,the enhanced intensity component and panchromatic image are integrated by weighted average method to reduce the lack of details in image fusion,and then the integrated image and panchromatic image are dealt with by histogram matching to adjust gray scale.Finally,the fused image is reconstructed by adopting inverse IHS transform.Experimental results indicate that spatial features of multispectral and panchromatic images are fully considered in the proposed algorithm.The clarity of fusion image is improved while preserving the spectral information.(2)An image fusion algorithm based on fractional differential and guided filtering is proposedIn order to resolve the issues of spectral distortion and lack of details in image fusion,an image fusion algorithm based on fractional differential and guided filtering is proposed in this paper.The basic idea is to use up-sampling operation into multispectral image firstly,then the multispectral image is transformed by IHS transform to obtain intensitycomponent.Subsequently,the guided filtering is employed in intensity component to reduce blocking artifacts of the multispectral image.Meanwhile,the panchromatic image is enhanced by adaptive fractional differential.Then the wavelet transform is utilized to fuse the enhanced panchromatic image and filtered intensity component.In the course of decomposition,the weighted average is applied in the low-frequency component.The high-frequency component is adopted with maximal principle.Finally,the fused image is obtained by inverse wavelet transform and inverse IHS transform.Experimental results on remote sensing images indicate that the proposed method achieves remarkable performance,which can preserve spectral information and provide rich spatial texture details.(3)An image fusion evaluation method based on structure tensor is proposedTo further explore the accuracy and reliability of the proposed algorithms,an image fusion evaluation method based on structure tensor is proposed in this paper.The fused image is divided into flat region,edge region,and corner region by calculating the trace of structure matrix in an image.For a remote sensing image,edge region and corner region are mainly considered because of their rich spatial information.Compared with the low-resolution remote sensing image,high-resolution remote sensing image is possessed of more edge and corner areas,and the number of trace points is richer in high-resolution remote sensing image.Thus,the number of trace points can be used to evaluate fusion quality.Experimental results indicate that the more data points mean the more detail features of fused image.Therefore,the image fusion evaluation method based on structure tensor is reliable and effective.