Research On Component Substitution Fusion Methods Of Remote Sensing Imagery

With the increasing ability of remote sensing satellite image data acquisition and sensor technology,a large number of high resolution remote sensing images have been sprung up.However,due to the limitation of physical conditions of sensors,it is difficult to directly acquire images with both high spatial resolution and high spectral resolution.Most of the earth observation satellites can only obtain high spatial resolution panchromatic and relatively low spatial resolution multispectral images.Remote sensing image fusion technology can integrate the advantages of these two images,and generate new high-quality images containing more information for scientific research and production.The existing fusion methods can be divided into three categories: ratio enhancement method,multi-resolution analysis method and component substitution method.Among them,the component substitution method is more efficient and practical,so it is more popular in practical applications.In addition,the generation of low resolution panchromatic bands and the design of weight injection model in the component substitution method are hot topics of the current research.This paper is aimed at the innovation of the fusion method of homologous remote sensing images in these two aspects.The main research contents and innovations of this paper are as follows:(1)The existing image fusion methods are classified,and some mature component substitution algorithms are introduced,such as IHS,PCA,GS,UNB,SRM and NND methods.Then,some representative direct evaluation and indirect evaluation methods for image fusion are summarized.Finally,based on the GaoFen-2 satellite images,some commonly-used component substitution image fusion methods are selected for comparison,the qualitative and quantitative evaluations are employed,and the limitations of the methods are summarized.(2)The guided filtering is intergated into the process of image fusion.A new approach based on guided filtering is proposed for image fusion.This method firstly uses the multispectral image to generate a low resolution panchromatic image.Then,the multispectral image is employed as a guidance image to the filtering process of lower resolution panchromatic image to extract the spatial information.Finally,inject the spatial details into the original multispectral image according to proper weights to obtain the fused images.The effectiveness of the method and the influence of parameters on the fusion effect are mainly analyzed and discussed.The PCA,GS,GSA,UNB and NND methods are applied to the experiments using GF-2,QuickBird and WorldView-2 satellite images respectively,and the fused results are directly and indirectly evaluated.Experimental results show that the ability of the proposed approach is superior to other traditional component substitution fusion methods in spatial detail enhancement and spectral information reservation,which can effectively improve the quality of the original images and increase the classification accuracy of images.(3)The general framework of component substitution image fusion is introduced in detail.And an improved fusion method based on the framework is proposed.Firstly,the approach uses the multispectral image to generate a low resolution panchromatic image.Then,extract the spatial details by subtracting the lower resolution panchromatic image from the original panchromatic image.Finally,the spatial details are injected into the original multispectral image according to proper weights,and thus the fused results are obtained.Several groups of remote sensing images acquired by different sensors and four typical component substitution fusion methods are employed for comparison.Subsequently,the fused images are evaluated using visual discrimination and quantitative evaluation,and the performance of the proposed method is analyzed and discussed in detail.The results show that the method can effectively preserve the spectral information of the original multispectral image while enhance the spatial details of the image,and it also can effectively improve the visual effect of the original image and meet the requirements of subsequent use of the image.