基于活动轮廓的序列图像分割方法研究

《基于活动轮廓的序列图像分割方法研究》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、西北工业大学硕士学位论文基于活动轮廓的序列图像分割方法研究姓名：蒋晓悦申请学位级别：硕士专业：计算机应用技术指导教师：赵忠明20020901摘要西北工业大学硕士学位论文AbstractThesegmentationofa目yenimageisakeystepfromtheImageProcessingtoImageUnderstanding．whichisalsoabasictechniqueintheComputerVisionItisforthereasonthattheimagecanchangetobemoreabstractandimpactedafterse

2、gmentingtheimage，separatingtheobjects，extractingthefeatures，measuringtheparametersandsoon．That'smadethehigh．1evelComputerVksiontaskspossible，suchasImageAnalysisandUnderstandingTheimagesegmentationhasbeenabsorbingmanyattentionsformanyyearsandagreatamountofthesegmentationalgorithmshavebeen

3、proposed．InthispaperasystemforsegmentingandtrackingⅡleobjectsfromSequencesisdesigned．Thissystemismainlymadeupofthreepartsthatarethemotionestimationgivingoutthemotionvectois，theedgedetectionfindingtheedgepointsandCol】tonrandthedefonnablecontourachievingthesegmentationandhackingfortheobjec

4、ts．ThroughtestingmanydifferentBlockMatchingAlgorithmsiIlthemotionestimation，theNewThreeStepSearchAlgorithmischosen．whichcanfulfilltherequirementofthesystem．ForingeneralthemotioninthesequencesissmallandtheNewThreeStepSearchisespeciallyeffectiveforcapturingthesmallmotion．Indesigningtheedge

5、detectionalgorithm，themulti·scalescharacterofthewaveletchangeandthecharacteroftheB—Splinefunctionasabestsmoothingfilterarecombined．BasedontheSchemeofLocalModulusMaxima，aB—Splinewaveletfunctionisdesignedtodetecttheedgepoints·Inthepost-processingfortheedgepoints，thecontourisfoundthroughthe

6、methodsbasedontheMathematicalMorphology．TheactivecontourischosentobethemethodforsegmentationThebasicideaoftheactivecontourisintroducedfirst．Acurvecanmovetothedestinationwiththeforcefromitselfandoutside，whenever1tsenergybecomeminimumitarrivestheaimpoints．Theintemalforcethatistheelasticity

7、andthestrengthmakesthecurceregularandsmooth．Theexternalforcegivenbvtheedgemapmakesthecnrvemovinginthedirectionoftheedge．1IltradifionaIactivecontourtheexternalforceisthepotentialenergyoftheedgemap，wInchhassoIlleinsurmountabledifficuriesThereforetheGVF(GradientVectorField)i