预览加载中,请您耐心等待几秒...
1/2
2/2

在线预览结束,喜欢就下载吧,查找使用更方便

如果您无法下载资料,请参考说明:

1、部分资料下载需要金币,请确保您的账户上有足够的金币

2、已购买过的文档,再次下载不重复扣费

3、资料包下载后请先用软件解压,在使用对应软件打开

Prewitt图像边缘检测方法的改进 Introduction: Prewittedgedetectionisacommonlyusedmethodforfindingedgesinanimage.However,likeotheredgedetectionalgorithms,itsuffersfromsomelimitations.Inthispaper,weproposemodificationstothePrewittedgedetectionmethodtoimproveitsaccuracyandovercomesomeofthechallengesitfaces. LimitationsofPrewittEdgeDetection: OneofthemainlimitationsofthePrewittedgedetectionmethodisthatitissensitivetonoise.Thismeansthatevensmallamountsofnoiseinanimagecanleadtofalseedgedetections.Additionally,thePrewittmethodhasdifficultydetectingedgesthatarenotperfectlyalignedwiththeverticalorhorizontaledges. AnotherdisadvantageofthePrewittedgedetectionmethodisthatitproducesthickedges.Thisisbecausetheoperatormeasuresthegradientmagnitudeonaneighborhoodofpixels.Asaresult,theedgepixelsareoftenlocatedatinconsistentpositions,resultinginanedgewithvariablethickness. ProposedModifications: TheproposedmodificationstothePrewittedgedetectionmethodhelptoovercomesomeofthelimitationsmentionedabove.OnemodificationinvolvestheuseofamedianfiltertoreducethenoiseintheimagebeforeapplyingthePrewittoperator.Thishelpstoeliminatefalseedgedetectionscausedbynoiseintheimageandenhancestheaccuracyoftheedgedetectionprocess. Anothermodificationistouseanadaptivethresholdingtechniquetodeterminetheedgepixels.Thismeansthatdifferentthresholdvaluesareappliedtodifferentregionsoftheimagebasedonthelocalimagestatistics.ThishelpstoovercomethelimitationofthePrewittmethodindetectingedgesthatarenotperfectlyalignedwiththeverticalorhorizontaledges. Toovercometheproblemofthickedges,weproposeamodificationthatinvolvesusinganon-maximumsuppressiontechnique.Thistechniqueinvolvesselectingonlythepixelsthathavethehighestgradientmagnitudeinalocalneighborhood.Thishelpstoreducethethicknessofthedetectededgesandleadstoamoreaccurateedgerepresentation. ExperimentalResults: Toevaluatetheeffectivenessoftheproposedmodifications,weconductedexperimentsonasetofbenchmarkimages.TheresultsshowedthatthemodifiedPrewittedgedetectionmethodoutperformedthetradi