应用Otsu改进Canny算子的图像边缘检测方法  

Application of Otsu to Improve Image Edge Detection Method in Canny Algorithm

作　　者：张志强[1,2] 宋海生[3] 

机构地区：[1]顺德职业技术学院,顺德528333 [2]华中科技大学计算机科学与技术学院,武汉430074 [3]中国科学院研究生院,北京100049

出　　处：《计算机与数字工程》2014年第1期122-128,141,共8页Computer & Digital Engineering

基　　金：国家自然科学基金(编号:41072247);广东省自然科学基金(编号:9151063201000065);顺德区科技计划项目(编号:20110202023);容桂镇2011年科技计划项目资助

摘　　要：针对传统Canny算子进行边缘检测时易丢失边缘细节的缺陷,提出了一种改进的Canny边缘检测算法。首先从数学形态学角度分析传统Canny思想和缺陷;接着提出应用尺度自适应调整的高斯滤波器改进传统高斯平滑滤波;然后使用最大类间分差法(Otsu)替代传统双阈值检测算法求出最佳阈值,有效平衡去噪能力和保留边缘细节信息二者之间的矛盾;为解决Otsu遍历时间长,实时性差的缺点,提出应用Kalman滤波器最小误差估计算法预先估计后续图像序列的阈值范围,以加快遍历过程;最后进行实验分析比较,证明该算法可以更快速有效地检测出图像的边缘。An improved Canny edge detection method is proposed due to the easily lost edge details when detected by the traditional Canny algorithr.The traditional Canny and its defects are analyzed in terms of mathematical morphology.Then the Gauss filter with scales for edge detection is applied to improve the traditional Gauss filter.The Otsu replaces the traditional dual—threshold algorithm to get the best threshold value which effectively balances the contradiction between the anti-noise function and the retaining of edge details.In order to solve the problems of the long-time ergodicity and the poor real-time capability,the minimum error estimation algorithm of Kalman filer is suggested to estimate the threshold range of subsequent image sequences in advance to facilitate the ergodicity process.Finally,through the experimental analysis,the algorithm is proved to detect the image edges more quickly and effectively.

关 键 词：CANNY 自适应滤波器 OTSU KALMAN 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]