摘要
SUSAN算子需要不断地调整"相似性阈值"才能获得更好的角点,它可以检测X型、Y型和T型,但是不能检测出特殊形状和复杂形状的角点。为了克服这些缺点,文中提出了一种自适应阈值的选择方法,使程序自动化,若角点的USAN区域与边缘点的USAN区域相同,则在SUSAN的模板中叠加一个离散的环状模板r,根据模板r中的亮度变化次数来区分待测点是否是角点,以免遗漏角点。实验结果显示改进后的算法不仅省去手动调整阈值,大大提高程序运行自动化,而且可以检测不同类型角点。
In order to find a better corner detection result,SUSAN operator needs to adjust the similarity threshold from time to time.It can detect most corners of X type,Y type,T type,but it can＇ t detect all special or complicated corners.In order to solve these problems,this paper comes up with a selection method of adaptive threshold to make the algorithm automation,if the corner＇ s USAN area is equal to the edge point,in case of missing corner,then adding a discrete ring-template r to SUSAN＇ s circle template,and according to the brightness change times of r to distinguish the point is corner or not.The experiment shows that the improved algorithm not only needn＇ t adjust the threshold,makes the algorithm automation,but can obtain a good detection of all shapes of corner.
出处
《信息技术》
2014年第1期31-34,共4页
Information Technology
基金
国家自然科学基金(51176129/E060703)
