基于改进Canny算法的噪声图像边缘检测  

Noise Image Edge Detection Based on Improved Canny Algorithm

作　　者：孙海明[1,2] 韩国强 Sun Haiming;Han Guoqiang(School of Mechanical Engineering,Hubei University of Automotive Technology,Shiyan 442002,China;Hubei Zhongcheng Technology Industry Technique Academy Co.Ltd,Shiyan 442002,China)

机构地区：[1]湖北汽车工业学院机械工程学院,湖北十堰442002 [2]湖北中程科技产业技术研究院有限公司,湖北十堰442002

出　　处：《湖北汽车工业学院学报》2023年第4期54-57,63,共5页Journal of Hubei University Of Automotive Technology

基　　金：湖北省揭榜制科技项目(2021BEC005)。

摘　　要：针对经典Canny算法应用中常出现的无法滤除椒盐噪声且滤波后图像细节信息易丢失、Sobel卷积核定位的边缘信息精度较差、双阈值选取存在偶然性等问题,对Canny算法进行改进。首先采用自适应中值-高斯滤波法代替传统的高斯滤波,并融合Laplace边缘增强法,滤除大量噪声的同时保留图像边缘细节信息;使用精度更高的Scharr算子代替Sobel算子计算图像梯度幅值和方向;然后通过最大类间方差法自适应计算图像的最优阈值;最后选用BSD500数据集进行实验,结果表明:文中算法相对于经典Canny算法,峰值信噪比平均提升14.5 dB,边缘检测评价指标C/A提高0.07~0.24,C/B提高0.06~0.14,算法性能指标提高24.8%。The Canny algorithm was improved to solve the problems often encountered in the applica⁃tion of the classical Canny algorithm,such as the inability to filter out salt-and-pepper noise,the loss of image details after filtering,the poor accuracy of edge information located by Sobel convolution ker⁃nel,and the contingency of double threshold selection.Firstly,the adaptive median Gaussian filtering method was used to replace traditional Gaussian filtering,and the Laplace edge enhancement method was fused to filter out a large amount of noise while preserving image edge details;higher-precision Scharr operator instead of Sobel operator was used to calculate image gradient amplitude and direction;then,the optimal threshold of the image was adaptively calculated by using the maximum inter-class variance method;finally,the BSD500 dataset was selected for experiments.The experimental results show that compared with the classic Canny algorithm,the proposed algorithm improves the peak signalto-noise ratio by 14.5 dB on average;the edge detection evaluation indexes C/A and C/B increase by 0.07-0.24 and 0.06-0.14,respectively,and the algorithm performance index improves by 24.8%.

关 键 词：边缘检测 CANNY算法 自适应去噪 自适应阈值 边缘增强 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]