Optical Remote Sensing Image Fusion Parameter Optimization And Quality Evaluation Method

With the development of hyperspectral remote sensing satellite technology,the spectral and spatial resolution of remote sensing images have been rapidly improved,even reaching sub-meter level,and the amount of information it contains tends to be complex and diverse.Among them,remote sensing image fusion is an important step in analyzing complex remote sensing information.Image fusion refers to a remote sensing technology that combines low spatial resolution multispectral images obtained by satellites with high spatial resolution single-band panchromatic images to obtain fused images with both high resolution and multispectral features.Image fusion algorithms can be divided into pixel-level image fusion,feature-level image fusion and decision-level image fusion according to different fusion levels.Because pixel-level image fusion algorithm has the advantages of high precision and fusion precision,this paper discusses remote sensing image fusion algorithm from the perspective of pixel-level image fusion and evaluation.Firstly,this paper briefly describes pixel-level image fusion algorithms,and analyzes the advantages and disadvantages,applicable scope and essence of pixel-level image fusion based on the experimental results of such algorithms.Secondly,the advantages and disadvantages of BRSVR image fusion algorithm in application are discussed and some improvements are made.Since the improved algorithm requires reasonable image segmentation,this paper discusses the influence of different segmentation scales on the fusion effect.Through the analysis of the principle and results,SLIC image segmentation algorithm is finally selected.Based on this,this paper designs an image fusion algorithm combining SLIC image segmentation and SVR method.Firstly,the image is divided into reasonable blocks by SLIC image segmentation method,and then the image is fused according to SVR image fusion method.Compared with other pixel-level image fusion algorithms,the SLIC-SVR algorithm proposed in this paper has better fusion results compared with other algorithms through objective quantitative evaluation based on six indexes such as standard deviation,entropy and normalized root mean square error.In addition,the fusion algorithm needs more scientific and convenient evaluation methods.On the basis of mutual information,this paper combines the advantages of qualitative and quantitative evaluation of remote sensing images,and constructs a histogram evaluation method of mutual information images.The method intuitively displays the difference of image spectral information in the way of gray-scale images,and is used for evaluating the spectral retention degree of fused images.Finally,the paper proves the application potential of SLIC-SVR remote sensing image fusion algorithm in image fusion through experiments and evaluation methods such as qualitative,quantitative and mutual information histograms.The main contributions of this paper are as follows:(1)Pansharpening image fusion algorithm is improved.combining the existing BRSVR image fusion theory and SLIC image segmentation algorithm,a SLIC-SVR image fusion algorithm is proposed.Combining the images of 7 satellites including GF-2,ZY-3,K3A and many complex ground object types,this paper compares and analyzes the experimental results of 6 traditional image fusion algorithms including wavelet transform,principal component analysis,IHS transform,and makes qualitative and quantitative evaluation.The experimental results show that SLIC-S VR algorithm has certain advantages in spatial detail injection and spectral feature preservation.(2)Based on information entropy and mutual information theory,combined with the form of gray statistical histogram,the image quality evaluation standard of mutual information histogram is proposed.This method can evaluate the image results qualitatively and quantitatively at the same time,which is a useful supplement to the image fusion quality evaluation index system.