基于Canny的边缘检测改进算法  

An Improved Edge Detection Method Based on Canny Algorithm

作　　者：李凌[1] LI Ling(Department of Computer Science and Technology,Huaibei Vocational&Technical College,Huaibei,Anhui235000,China)

机构地区：[1]淮北职业技术学院计算机科学技术系

出　　处：《通化师范学院学报》2019年第6期61-64,77,共5页Journal of Tonghua Normal University

基　　金：安徽省自然科学重点研究项目(KJ2017A525);安徽省高校学科拔尖人才学术资助重点项目(gxbjZD2016113)

摘　　要：传统Canny边缘检测算法的效果易受噪声、光照等因素的影响,且需人工选择高斯滤波方差与滞后阈值,自适应性差.提出改进的Canny边缘检测算法,该算法采用复合形态滤波取代高斯滤波,降低噪声的影响,保持边缘强度与细节;采用加入了45°和135°方向模板的Sobel算子计算梯度幅值,提高边缘定位效果;根据图像的灰度均值和方差选取阈值,边缘检测更连续.实验表明:改进的算法比传统Canny算法具有更好的检测效果和抗噪性能.The traditional Canny algorithm without the adaptive ability in the variance of Gaussion filtering and threshold,is susceptibale to noise,light and so on. This paper proposes an improved edge detection method based on Canny alogorithm. Firstly,compound morphology smoothing replaces Gaussian filtering,not only reduces the influence of noise,but also maintains the edge strength and details. Then,the gradient amplitude is calculated by using the Sobel detector with adding 450 and 1350 direction templates,so as to improve the effect of edge location. Finally,the thresholds is automatically seleced according to the mean of variance and gray-scale,so that the edge detection is more continuous. Experiment shows the improved algorithm has a good effect and anti-noise performance.

关 键 词：边缘检测 滤波 梯度幅值 CANNY算子 SOBEL算子 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]