ItiswellknownthatLandsatTMimagesarethemostwidelyusedremotesensingdatainvariousfields.Usually,ithas7differentelectromagneticspectrumbands,amongwhichthesixthonehasmuchlowergroundresolutioncomparedwiththeothersixbands.Nevertheless,itisusefulinthestudyofrockspectrumreflection,geo-thermalresourcesexploration,etc.ToimprovethegroundresolutionofTM6tothelevelasthatoftheothersixbandsisaproblem.Thispaperpresentsanalgorithmbasedonthecombinationofmulti-variateregressionmodelwithsemi-variogramfunctionwhichcanimprovethegroundresolutionofTM6by"fusing"thedataofothersixbands.Itincludesthefollowingmainsteps:(1)testingthecorrelationbetweenTM6andoneofTM1-5,7.IfthecorrelationcoefficientbetweenTM6andanotheroneisgreaterthanagiventhresholdvalue,thenselectthebandtotheregressionanalysisasanargument.(2)calculatingthesizeofthetemplatewindowwithinwhichsomeparametersneededbytheregressionmodelwillbecalculated;(3)replacingtheoriginalpixelvaluesofTM6bythoseobtainedbyregressionanalysis;(4)usingimageentropyasameasurementtoevaluatethequalityofthefusedimageofTM6.ThebasicmechanismofthealgorithmisdiscussedandtheVC++programforimplementingthisalgorithmisalsopresented.Asimpleapplicationexampleisgiveninthelastpartofthispaper,showingtheeffectivenessofthealgorithm.
