欢迎来到天天文库
浏览记录
ID:40643819
大小:80.00 KB
页数:8页
时间:2019-08-05
《外文文献资料(Canny算子)》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、苏州大学本科生毕业设计(论文)附件:外文文献资料与中文翻译稿外文文献资料Cannyedgedetector1.CannyedgedetectorIntroductionTheCannyedgedetectionoperatorwasdevelopedbyJohnF.Cannyin1986andusesamulti-stagealgorithmtodetectawiderangeofedgesinimages.Mostimportantly,Cannyalsoproducedacomputationaltheoryofedgedetectionexplainingwhythetechniq
2、ueworks.2.DevelopmentoftheCannyalgorithmCanny'saimwastodiscovertheoptimaledgedetectionalgorithm.Inthissituation,an"optimal"edgedetectormeans:lgooddetection–thealgorithmshouldmarkasmanyrealedgesintheimageaspossible.lgoodlocalization–edgesmarkedshouldbeascloseaspossibletotheedgeintherealimage.lmini
3、malresponse–agivenedgeintheimageshouldonlybemarkedonce,andwherepossible,imagenoiseshouldnotcreatefalseedges.TosatisfytheserequirementsCannyusedthecalculusofvariations–atechniquewhichfindsthefunctionwhichoptimizesagivenfunctional.The8苏州大学本科生毕业设计(论文)附件:外文文献资料与中文翻译稿optimalfunctioninCanny'sdetectoris
4、describedbythesumoffourexponentialterms,butcanbeapproximatedbythefirstderivativeofaGaussian.3.StagesoftheCannyalgorithm3.1NoisereductionTheCannyedgedetectorusesafilterbasedonthefirstderivativeofaGaussian,becauseitissusceptibletonoisepresentonrawunprocessedimagedata,sotobeginwith,therawimageisconv
5、olvedwithaGaussianfilter.Theresultisaslightlyblurredversionoftheoriginalwhichisnotaffectedbyasinglenoisypixeltoanysignificantdegree.3.2FindingtheintensitygradientoftheimageAnedgeinanimagemaypointinavarietyofdirections,sotheCannyalgorithmusesfourfilterstodetecthorizontal,verticalanddiagonaledgesin
6、theblurredimage.Theedgedetectionoperator(Roberts,Prewitt,Sobelforexample)returnsavalueforthefirstderivativeinthehorizontaldirection(Gy)andtheverticaldirection(Gx).Fromthistheedgegradientanddirectioncanbedetermined.3.3TracingedgesthroughtheimageandhysteresisthresholdingIntensitygradientswhicharela
7、rgearemorelikelytocorrespondtoedgesthaniftheyaresmall.Itisinmostcasesimpossibletospecifyathresholdatwhicha8苏州大学本科生毕业设计(论文)附件:外文文献资料与中文翻译稿givenintensitygradientswitchesfromcorrespondingtoanedgeintonotdoingso.ThereforeCa
此文档下载收益归作者所有