自适应多窗口梯度幅值边缘检测算法  

Edge-detection algorithm based on adaptive multi-window gradient magnitude

作　　者：刘勇[1] 姚刚[1] 肖人彬[1,2] 雷帮军[1] 

机构地区：[1]三峡大学计算机与信息学院,湖北宜昌443002 [2]华中科技大学控制科学与工程系,湖北武汉430074

出　　处：《华中科技大学学报（自然科学版）》2011年第1期14-18,共5页Journal of Huazhong University of Science and Technology(Natural Science Edition)

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

摘　　要：针对梯度幅值边缘检测算法无法检测连续边缘的问题,提出一种自适应多窗口梯度幅值边缘检测算法.首先使用传统梯度幅值边缘检测算法检测出初始边缘;然后在初始边缘上检测端点,使用K-均值聚类算法对端点进行分类,从而确定背景和目标灰度值接近的区域作为窗口;最后在窗口内使用梯度幅值检测边缘,通过多个窗口的并集得到最终的边缘.实验证明采用所提出的算法可以得到比较完整的边缘图,定位误差比传统的边缘检测算法小.Aiming at the defects of the traditional edge-detection algorithms based on gradient magnitude,which lead to some discontinuous edge,an edge-detection algorithm based on adaptive multi-windows gradient magnitude was proposed.Firstly,traditional gradient magnitude algorithm was adopted to detect the original edge,and then the endpoints in the original edge image were detected.Next,the K-means clustering algorithm was used to classify the endpoints.Finally,the multiple windows using adaptive gradient magnitude were merged to get the final edge map.The experimental results show that the algorithm can get more complete edge map with smaller position error by comparing with the traditional edge detection algorithms.

关 键 词：梯度幅值 边缘检测 自适应 多窗口 K-均值聚类 端点检测 

分 类 号：TN911.73[电子电信—通信与信息系统]