Opto-ElectronicEngineering2009年2月Feb,2009文章编号：1003-501X(2009)02-0091-05ANewRemoteSensingImageFusionAlgorithmintheDecimatedWaveletDomainCHUHeng1，CHENHua-gang1，ZHUWei-le2(1.ChongqingSurveyInstitute,Chongqing400020,China;2.SchoolofElectronicEngineering,UniversityofElectronicScienceandTechnologyofChina,Chengdu610054,China)Abstract:Inviewofthefaultsofconventionalimagefusionmethods,anovelremotesensingimagefusionalgorithmwithhighspatialqualitybasedonIHStransformandthedecimatedwavelettransformisproposed.Themergedhighfrequencycoefficientscomefromthehighfrequencycoefficientsofthehistogrammatchedpanchromaticimagewiththeproposedfusionmethod,whilethefusedlowfrequencycoefficientsareobtainedaccordingtothefusionrulesuggestedinthispaper.ThemergedimageisacquiredwiththeinversewavelettransformandtheinverseIHStransform.Experimentalresultsshowthattheproposedalgorithmprovidesabettertradeoffbetweenimprovingspatialresolutionandpreservingspectralquality.Thecomputationalcomplexityoftheproposedmethodisclosetothatoftheconventionalfusionapproachbasedondecimatedwavelettransformandmuchlowerthanthatofthefusionmethodbasedonundecimatedwavelettransform.Keywords:imagefusion;remotesensing;IHStransform;wavelettransformCLCnumber:TN911.73Documentcode:A一种抽样小波域的遥感影像融合新方法楚恒1，陈华刚1，朱维乐2(1.重庆市勘测院，重庆400020；2.电子科技大学电子工程学院，成都610054)摘要：针对传统图像融合方法的不足，提出了一种新的基于IHS变换与抽样小波变换的高清晰遥感影像融合方法。该方法将直方图匹配后的全色影像的高频系数作为融合后的高频系数，而融合后的低频系数依据新提出的融合规则得到，最后采用小波逆变换与逆IHS变换获得融合图像。实验结果表明，该方法在提高融合影像的空间分辨率与光谱质量之间得到了更好的折衷。该方法的计算复杂度接近于常用的基于抽样小波变换的融合方法，明显少于基于非抽样小波变换的融合方法。关键词：图像融合；遥感；IHS变换；小波变换0IntroductionImagefusion,alsoknownasimagemergingorpansharpening,isatechniqueemployedtomergethespatialdetailsofahighspatial-resolutionpanchromatic(Pan)imageandthecolorinformationofalowspatial-resolutionmultispectral(MS)imagetoproduceahighspatial-resolutionMSimage[1-3].Animagewithhighspatialandspectralresolutionisnecessaryformanyremotesensingapplications,especiallyGIS-basedapplications[2].ConventionalimagefusionmethodsincludetheIntensity-Hue-Saturation(IHS)transform,principalcomponentanalysis(PCA),andBroveytransform(BT)[1],whichareadoptedbymanycommercialsoftwaresfor收稿日期：2008-07-22；收到修改稿日期：2008-09-25作者简介：楚恒(1976-)，男(汉族)，四川渠县人。博士，主要研究工作是图像处理，小波变换等。E-mail:chuheng2000@163.com。Opto-ElectronicEngineering2009年2月92remotesensingimagefusion.However,theseapproachesshowhighcolordistortionofthespectralinformation,especiallyfortheremotesensingimagesofIKONOSandQuickBird[1-2].Multiresolutionanalysis(MRA)basedimagefusionalgorithmshavebeenrecognizedwidelyasoneofthemostefficienttoolsforremotesensingimagefusion[2-7].However,thedecimatedDWT(DiscreteWaveletTransform)basedimagefusionschemesoftencausetheringingartifactsinthefusedimagesforitscriticaldownsampling.AlthoughanundecimatedDWT,whichisashift-invariantversionoftheDWT,canavoidsomeoftheartifacts,thecomputationalcomplexityoftheundecimatedDWTincreaseswiththenumberofdecompositionlevels.Thecomputationalcostofthecurvelet-basedmethodproposedbyNencinietal.[6]isalsoveryhigh.HaohaoSongetal.suggestedaco...