lec4 Intensity Transformations and Spatial Filtering (II)

《lec4 Intensity Transformations and Spatial Filtering (II)》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、ImageAnalysisandComputerVisionLecture4!ImageEnhencementinSpatialFiltering(II)WeiqiangWangSchoolofInformationScienceandEngineering,GUCASSeptember19,2010W.Q.Wang(SISE,GUCAS)ImageAnalysisandComputerVisionSeptember19,20101/44Outline1HistogramProcessingandFunc

2、tionPlotting2SpatialFiltering3StandardSpatialFiltersinImageProcessingToolboxW.Q.Wang(SISE,GUCAS)ImageAnalysisandComputerVisionSeptember19,20102/44Outline1HistogramProcessingandFunctionPlotting2SpatialFiltering3StandardSpatialFiltersinImageProcessingToolbo

3、xW.Q.Wang(SISE,GUCAS)ImageAnalysisandComputerVisionSeptember19,20102/44Outline1HistogramProcessingandFunctionPlotting2SpatialFiltering3StandardSpatialFiltersinImageProcessingToolboxW.Q.Wang(SISE,GUCAS)ImageAnalysisandComputerVisionSeptember19,20102/44Hist

4、ogramProcessingandFunctionPlottingHistogramMatchingHistogrammatchingissimilartohistogramequalization,exceptthatinsteadoftryingtomaketheoutputimagehavea athistogram,wewouldlikeittohaveahistogramofaspeciedshape.Considerforamomentcontinuouslevelsthatarenorm

5、alizedtotheinterval[0,1],andletrandzdenotetheintensitylevelsoftheinputandoutputimages.Theinputlevelshaveprobabilitydensityfunctionpr(r)andtheoutputlevelshavethespeciedprobabilitydensityfunctionpz(z).W.Q.Wang(SISE,GUCAS)ImageAnalysisandComputerVisionSepte

6、mber19,20103/44HistogramProcessingandFunctionPlottingHistogramMatchingWeknowfromthediscussionintheprevioussectionthatthetransformation:Zrs=T(r)=pr(w)dw0resultinaidealequalizedhistogramps(s).SupposenowwedeneavariablezwiththepropertyZzH(z)=pz(w)dw=s0Fromth

7、eprecedingtwoequations,itfollowsthatz=H1(s)=H1(T(r));WecanndT(r)fromtheinputimage(thisisthehistogramequalizationtransformation),soitfollowsthatwecanusetheprecedingequationtondthetransformedlevelszwhosePDFisthespeciedpz(z)aslongaswecanndH1.W.Q.Wang(

8、SISE,GUCAS)ImageAnalysisandComputerVisionSeptember19,20104/44HistogramProcessingandFunctionPlottingHistogramMatchingThetoolboximplementshistogrammatchingusingthefollowingsyntaxinhisteq:g=histeq(f,hspec)wherefisthein