基于Matlab的数字图像边缘检测算法研究

沈阳大学毕业设计（论文）基于Matlab的数字图像边缘检测算法研究专业：自动化班级：2006级4班姓名：沈阳大学毕业设计（论文）目录引言................................................................................................................................................31基于一阶微分的边缘检测算法.....................................................................................71.1Roberts算子.....................................................................................................................71.2Prewitt算子......................................................................................................................81.3Kirsch算子.......................................................................................................................82基于二阶微分的边缘检测算法...................................................................................102.1Laplacian算子................................................................................................................102.2LoG算子........................................................................................................................122.3Canny算子.....................................................................................................................153最佳Sobel边缘检测方法..............................................................................................183.1Sobel图像边缘检测方法...............................................................................................183.2改进的Sobel边缘检测方法..........................................................................................183.3Sobel算子的最佳阈值选取...........................................................................................194基于Matlab的实验结果与分析..................................................................................224.1Matlab简介....................................................................................................................224.2一阶微分算法的实验结果与分析.................................................................................254.3二阶微分算法的实验结果与分析.................................................................................304.4最佳Sobel边缘检测算法的实验结果与分析..............................................................35结论..............................................................................................................................................40致谢..............................................................................................................................................42参考文献....................................................................................................................................43附录源程序清单..................................................................................................................45沈阳大学毕业设计（论文）No1摘要在实际图像处理问题中，图像的边缘作为图像的一种基本特征，经常被应用到较高层次的图像应用中去。它在图像识别，图像分割，图像增强以及图像压缩等的领域中有较为广泛的应用，也是它们的基础。边缘检测是图像处理与分析中最基础的内容之一，也是至今仍没有得到圆满解决的一类问题。图像的边缘包含了图像的位置、轮廓等特征，是图像的基本特征之一，广泛地应用于特征描述、图像分割、图像增强、图像复原、模式识别、图像压缩等图像分析和处理中。因此，图像边缘和轮廓特征的检测与提取方法，一直是图像处理与分析技术中的研究热点，新理论、新方法不断涌现。本文研究了一些边缘检测算法，包括传统的Roberts、Sobel、Prewitt、LoG、Canny、Kirsch等算法。经典边缘检测方法的抗噪声性能都较差，解决该问题的主要方法就是设置阈值，把得到的图像高频部分与阈值相比较以达到去噪的目的，所以阈值的选取显得尤为重要。传统方法中的阈值都是通过实验确定的，没有统一的阈值选取方法。本文利用边缘的最大后验概率估计，介绍一种新的边缘估计方法，从理论上说明了怎样选取最佳阈值。文章中关于这些方法都有较详细的介绍，以及算法的实现步骤。对算法均进行了仿真实验。论文的主要目的是进行图像边缘检测算法性能比较的研究。实验结果表明，本文的算法比较可以为图像处理的后续环节提供一些有益的参考。关键词：边缘检测；图像处理；Matlab；Sobel；检测算法沈阳大学毕业设计（论文）No2AbstractInimageprocessing,asabasiccharacteristic,theedgeoftheimage,whichiswidelyusedintherecognition,segmentation,intensificationandcompressoftheimage,isoftenappliedtohigh-leveldomain.Edgedetectionisoneofthemostfundamentalsinimageprocessingandanalyzing,whichisstillunsolved.Image’sedgesincludeimage’sfeaturessuchaspositionandoutline,whichbelongtothefundamentalfeatures.Edgedetectioniswidelyusedinimageanalysisandprocessingsuchasfeaturedescription,imagesegmentation,imageenhancement,patternrecognitionandimagecompressionetc,soEdgeDetectionandextractofoutlinefigurearetheresearchhotspotinthetechnologyofimageprocessingandanalysisall,onwhichthenewtheoryandmethodsareputforwardconstantly.Someedgedetectionalgorithmsareresearched,includingRoberts,Sobel,Prewitt,LoG,Canny,Kirschetc.Theclassicedgedetectionmethodshavesomedrawbacksinimagedenoising.Themainideaofresolvingthisproblemistosetathreshold,thencomparethethresholdwiththehigh-frequencycomponentsofanimagetoremovethenoise.Thechoiceofthethresholdisthekeyofimagedenoising.Thethresholdoftraditionalmethodsisobtainedbyexperiments,andthereisnogeneralmethodtodetermineit.AnewmethodofedgeestimationbytheMaximumaPosteriori(MAP)ispresentedinthispaper.Itproveshowtochoosetheoptimalthresholdintheory.Thesealgorithmsareintroducedindetail,andthestepsofalgorithmrealizationareintroducetoo.Thealgorithmistestedbysimulation.ThemainaimofthepaperisstudyingPerformanceEvaluationinedgedetectionalgorithms,andthedetectionresultsarecompared.Theexperimentsresultsshowthatthealgorithmscanbeausefulreferenceoflatterpartsofimageprocessing.Keywords:edgedetection;imageprocessing;Matlab;Sobel;detectionalgorithm沈阳大学毕业设计（论文）No3引言图像边缘是图像最基本的特征