该评
价函数描述了类间方差;最后依次搜索评价函数的一个最大值,自动获取
Canny
算子的高低阈值。该方法不需要
人为设定任何参数。
与传统
Canny
算法、
直接
Otsu
算法的结果比较表明,
本文方法对不同的图像都能较好地提取
真实边缘,特别是在低对比度图像的边缘提取上,此方法更具有优势。
关键词:
边缘检测;
Canny
算法;改进的
Otsu
算法;自适应阈值
中图分类号:
TP391.41
文献标志码:
A
doi:
10.3969/j.issn.1003-501X..05.023
AnImprovedAlgorithmforCannyEdge
DetectionwithAdaptiveThreshold
TANGLu-lu
1
,
ZHANGQi-can
1
,
HUSong
2
(1.
SchoolofElectronicandInformationEngineering
,
SichuanUniversity
,
Chengdu
610065,
China
;
2.
InstituteofOpticsandElectronics
,
ChineseAcademyofSciences
,
Chengdu
610209,
China
)
Abstract:
ToovercomethedifficultyofthresholdselectinginCannyalgorithm,animprovedmethodbasedonOtsu
algorithmisproposedtochoosethethresholdadaptivelyandsimultaneously.Firstly,guidedbythegradienthistogramof
thetestimage,allthepixelsaredividedintothreeclasses.Secondly,basedontheimprovedOtsualgorithm,anevaluation
functionisdefinedtodescribethemeansquareerroramongthethreeclasses.Finally,boththehighandlowthresholds
areselectedadaptivelyandindependentlybysearchingthemaximumvaluesoftheevaluationfunction.Artificial
paredwiththeresultsfromtraditionalCannymethodandDirect
Otsumethod,themethodshowsgreatadvantageinextractingtherealedgesfromdifferentimages,especiallylow
contrastones.
Keywords:
edgedetection;Cannyalgorithm;improvedOtsualgorithm;adaptivethreshold
0引言
随着当前数字化信息技术的发展,数字图像处理技术变得越来越重要,尤其在光电探测领域更是得到