基于多尺度分析的多光谱与全色图像融合技术分析-analysis of multispectral and panchromatic image fusion technology based on multiscale analysis.docxVIP

AbstractItisdesirabletohavetheimageswithbothhighspectralandspatialresolutionforphotoanalysisandimprovingtheresultsofanumberofapplicationslikelandcoverclassification,changemonitoring,andobjectdetection.Theprocessofpansharpeningaimsatincreasingthespatialresolutionofmultispectral(MS)imagesaswellaspreservingtheirspectralsignaturebythehelpofthespatialinformationofhigh-resolutionpanchromatic(Pan)images.Recently,afamilyofmultiresolutiontransformsnamedX-letsbasedfusionmethodshavebeenproposedtoovercomethelimitationofthewaveletfromtheperspectiveoftheexpressionofdirectionalinformationandintrinsicgeometricalstructures.Theaimofthispaperistoassesscurvelet,contourlet,NSCTandshearletbasedfusionmethodsandtoinvestigatethatwhichoneisbettersuitedforpansharpening.Firstly,thedefinitionofpansharpeningisdiscussed.ThetheoryofX-letstransformsandARSISconceptisreviewed.FourmultiscalemodelsbasedonX-letstransformsareconstructedtofulfillthefusiontask,whilethreegeneralMRA-basedimagefusionalgorithmsarepresented.TwoofthemaredevelopedundertheARSISconcept.AnotheronecanbelabeledasahybridmodelincorporatingbothPCAandX-lets.Inaddition,theindexQ4isgeneralizedtoQ8for8-bandWorldView2(WV2)imageusingthetheoryofoctonions.BoththeoreticalanalysisonextensibilityofindexQandexperimentalresultsprovetheproposedindexQ8inheritsthesubtlediscriminationcapabilityofspectraldistortionandofinaccuraciesinspatialenhancementfromQ4,successfully.Inthelast,athoroughcomparisonamongcurvelet,contourlet,nonsubsampledcontourlettransform(NSCT)andshearlettransformbasedfusionmethodsisperformedinthreedifferentinjectionmodel:ARSISglobalmodel,ARSISlocalmodelandhybridmodel.VisualandobjectivecomparisonsarepresentedonQuickBird(QB)andWV2imagedata,aswellastherobustassessment.TheresultsshowsthatNSCTandshearlettransformoutperformtheothers.Concerningthecomputationcomplexity,shearlettransformisrecommendedforpansharpening.Keywords:imagefusion;multiscaletransform;X-letstransform;ARSISconcept;qualityassessment目录摘要.........................................