应用Otsu改进Canny算子的图像边缘检测方法

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.

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

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.