图像分割概述.ppt
《图像分割概述.ppt》由会员分享,可在线阅读,更多相关《图像分割概述.ppt(21页珍藏版)》请在冰豆网上搜索。
图像分割图像分割图像识别与人工智能研究所图像识别与人工智能研究所,多谱信息处理国家重点实验室多谱信息处理国家重点实验室陶文兵陶文兵华中科技大学图像识别与人工智能研究所多谱信息处理技术国家重点实验室分割的目的和意义w图像分割是计算机视觉研究中的基础问题和难点之一w图像分割就是把图像分成各具特性的区域并提取出感兴趣目标w图像分割的难点和挑战性n对一般图像中的大量视觉模型进行建模的复杂性n图像理解本身的内在模糊性n当没有一个明确的任务来指导注意机制2图像工程的三层模型imagesegmentationGoal:
BreakuptheimageintomeaningfulorperceptuallysimilarregionsTypesofsegmentationsOversegmentationUndersegmentationMultipleSegmentationsSegmentationasaresultRotheretal.2004SegmentationforefficiencyFelzenszwalbandHuttenlocher2004Hoiemetal.2005,Mori2005ShiandMalik2001SegmentsasprimitivesforrecognitionJ.TigheandS.Lazebnik,submittedtoECCV2010MajorprocessesforsegmentationwBottom-up:
grouptokenswithsimilarfeatureswTop-down:
grouptokensthatlikelybelongtothesameobjectLevinandWeiss2006Bottom-upsegmentationGrouptogethersimilar-lookingpixelsforefficiencyoffurtherprocessingn“Bottom-up”processnUnsupervisedX.RenandJ.Malik.Learningaclassificationmodelforsegmentation.ICCV2003.“superpixels”ThegoalsofsegmentationSeparateimageintocoherent“objects”n“Bottom-up”or“top-down”process?
nSupervisedorunsupervised?
Berkeleysegmentationdatabase:
http:
/www.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/segbench/imagehumansegmentationTop-downsegmentationwE.BorensteinandS.Ullman,“Class-specific,top-downsegmentation,”ECCV2002wA.LevinandY.Weiss,“LearningtoCombineBottom-UpandTop-DownSegmentation,”ECCV2006.Top-downsegmentationwE.BorensteinandS.Ullman,“Class-specific,top-downsegmentation,”ECCV2002wA.LevinandY.Weiss,“LearningtoCombineBottom-UpandTop-DownSegmentation,”ECCV2006.NormalizedcutsTop-downsegmentation图像分割方法的发展现状w目前图像分割方法主要有三个比较重要的方向:
n基于统计理论的图像分割方法MeanShift,DDMCMCn基于变分模型的图像分割方法SnakeModel,GAC,M-SModel,C-VModeln基于图论的图像分割方法GraphCuts,NormalizeCuts14ThreebasictheoryinImageSegmentationStatisticsVariationalGraphTwobasicModelinImageSegmentationStatisticsformulationVariationalModelEnergyModelsOptimizationMethodBayesianformulationBayesianformulation(GemanandGeman,1984)Geman,S.andD.Geman:
1984,“Stochasticrelaxation,Gibbsdistributions,andtheBayesianrestorationofimages”.IEEETransactionsonPatternAnalysisandMachineIntelligence6,721741.(13641)Dpisadatapenaltyfunction,Vp,qisaninteractivefunctionDatapenaltiesindicateindividuallabel-preferencesofpixelsbasedonobservedintensitiesandprespecifiedlikelihoodfunction.Interactionfunctionencouragespatialcoherencebypenalizingdiscontinuitiesbetweenneighboringpixels.MAP-MRF:
Maximumaposteriori-MarkovrandomfieldSnakemodel(Kassetal.,1988)Kass,M.,A.Witkin,andD.Terzopoulos:
1988,“Snakes:
Activecontourmodels”.InternationalJournalofComputerVision,vol.1,pp.321331(13622)SnakemodelInternalenergyThefirsttwotermscontrolthesmoothnessofthecontourstobedetected.ExternalenergyThethirdtermisresponsibleforattractingthecontourtowardstheobjectintheimage.Geodesicactivecontoursmodel(GAC)(Casellesetal.,1997)V.Caselles,R.Kimmel,andG.Sapiro,“Geodesicactivecontours,”Int.J.Comput.Vis.,vol.22,pp.6179,1997.(3709)Remarks:
Thefunctiongisanedgeindicatorfunctionthatvanishesatobject.Theshorterthecurve,thelargerthegradientofthecurve,thelesstheenergy.Edge-basedActiveContourGeodesicactivecontoursmodelMumfordandShahfunctional(MumfordandShah,1989)Mumford,D.andJ.Shah:
1989,Optimalapproximationsbypiecewisesmoothfunctionsandassociatedvariationalproblems.CommunicationsonPureandAppliedMathematics42,577685.(3122)Mumford-ShahfunctionalRemarks:
TheminimizationofMumford-ShahfunctionalresultsinanoptimalcontourCthatsegmentsthegivenimageu0intoseveralregions.Imageuisanoptimalpiecewisesmoothapproximationofthegivenimageu0ImageuissmoothwithineachoftheconnectedcomponentsintheimagedomainseparatedbythecontourC.CVModel-piecewiseconstantMSmodel(ChanandVese,2001)T.ChanandL.Vese,“Activecontourswithoutedges,”IEEETrans.ImageProcess.,vol.10,no.2,pp.266277,Feb.2001.Citedtimes:
4514Region-basedActiveContourRemarks:
Assumethatuisapiecewiseconstantfunction.Forsuchcase,thesecondtermdisappearsfromtheMSfunctions.Twophaseproblem,c1istheaverageofregion1,c2istheaverageofregion2.Chan-Vese(CV)modelOptimizationMethod1、模拟退火(simulatedannealing)2、水平集算法(LevelSet)3、图割算法(Graphcuts)4、期望最大化算法(Expectation-MaximizationEM)5、置信传播(Beliefpropagation)4、对偶算法(primaldual)