一种自适应梯度幅值的边缘检测算法  

A Edge-detection Algorithm Based on Adaptable Gradient Magnitude

作　　者：姚刚[1] 刘勇[1] 夏道平[1] 王勇[1] 

机构地区：[1]三峡大学智能视觉与图像信息研究所,宜昌443002

出　　处：《计算机与数字工程》2010年第11期17-21,共5页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:50805087;60972162);湖北省自然科学基金(编号:2009CDZ027);三峡大学研究生创新基金项目(编号:200914)资助

摘　　要：提出了一种基于Otsu自适应梯度幅值的边缘检测算法。鉴于传统基于梯度幅值的自适应边缘检测算法中单阈值所带来的边缘断裂问题,分别使用传统Otsu算法和迭代Otsu算法进行自适应阈值,能保证图像梯度幅值的方差最大的情况下求得最优阈值,从而准确的检测边缘像素;针对梯度幅值边缘检测算法中边缘较宽的问题,在Otsu自适应梯度幅值检测算法检测到的边缘图基础上,使用经典的并行细化算法进行了边缘细化;最后,通过实验结果与对比分析,验证了算法能准确定位边缘和获得完整的边缘信息,实现了较好的边缘检测效果。An adaptable gradient magnitude edge-detection algorithm based on Otsu is proposed in this paper.In view of the defects of adaptive edge-detection algorithm based on gradient magnitude,in which the single threshold bring out the problem of marginal fracture.The adaptive threshold value is calculated basing on the traditional Otsu algorithm and Iterative Otsu algorithm,in the case of to ensure the variance of image magnitude of the biggest and then get the optimal threshold to accurately detect the edge pixels,and then,on the bases of edge image which get from adaptive Otsu gradient magnitude,classical parallel thinning algorithm are adopted to the issue of wide edge.Finally,by means of some experimental results and analysis,the advantages of accurately locate the edge and get the complete edge information and edge detection algorithm are demonstrated.

关 键 词：梯度幅值 OTSU 边缘检测 

分 类 号：TN911.73[电子电信—通信与信息系统]