*EqualizedContrastStretchThisstretchassignsmoredisplayvalues(range)tothefrequentlyoccurringportionsofthehistogram.Inthisway,thedetailintheseareaswillbebetterenhancedrelativetothoseareasoftheoriginalhistogramwherevaluesoccurlessfrequently*EqualizedContrastStretch*SpatialFiltering?Toimproveinterpretabilityofimagedata?Toaidinautomatedfeatureextraction?Toremoveand/orreducesensordegradation**Low-passFilterLinearStretchedImageLow-passFilterImage*High-passFilters:LinearContrastStretchHi-passFilter***DirectionalFilters??????????????????????????????????EdgeDetection
LakesStreamsEdgeDetection
FracturesShorelineenhancelinearfeatures**DensitySlicingGOESSatelliteinfrareddataissubdividedinto6colorlevelsfromcold(white,gray,purple)tocoldest(brown,red,darkbrown).*InformationExtractionFalseColorCompositesImageRatios,PrincipleComponentsAnalysis*FalseColorCompositesBand1Band2Band3*ImageRatios:SensorImageRatioEMSpectrumApplicationLandsatTMBands3/2red/greenSoilsLandsatTMBands4/3PhotoIR/redBiomassLandsatTMBands7/5SWIR/NIRClayMinerals/RockAlteration?NormalizedDifferenceVegetationIndex(NDVI):NDVI=100*SquareRoot[(IR-Red)/(IR+Red)]*PrincipleComponentsAnalysis:*PCAImageExample:Falsecolorimagewithcontraststretch(SPOTdata)PCAdecorrelationstretchofsameimage*UnsupervisedClassificationsSupervisedClassifications*UnsupervisedClassifications:*K-meansK均值算法的聚类准则是使每一聚类中，多模式点到该类别的中心的距离的平方和最小。其基本思想是：通过迭代，逐次移动各类的中心，直至得到最好的聚类结果为止。**SupervisedClassification:*MinimumDistance方法:最小距离法:在利用训练数据获得了各个分类类别的特征参数(如均值向量)以后,对于一未知像元,首先计算它与各个类别特征向量或代表特征向量(如均值向量)的距离,然后比较距离的大小,未知像元归并到距离最小的类别中.**Electro-magneticradiationReflected,Emittedsensor(remotesensor)Cameras,scannersPlatformAircraft,satellitesDataProcessinganditsApplication*2.Use