Prewitt图像边缘检测方法的改进-豆柴文库

您所在位置：网站首页 / Prewitt图像边缘检测方法的改进.docx / 文档详情

在线预览结束，喜欢就下载吧，查找使用更方便

5 金币

下载文档

/ 2

下载提示

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

文本预览

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