基于多级滤波的感兴趣边缘检测算法  

Algorithm of interesting edge detection based on multi-level filter

作　　者：王明静[1] 张学峰[1] 

机构地区：[1]中国空空导弹研究院,河南洛阳471009

出　　处：《激光与红外》2014年第3期335-338,共4页Laser & Infrared

摘　　要：针对边缘检测过程中阈值难确定和提取干扰边缘较多的问题,提出了一种基于多级滤波的感兴趣边缘检测算法。该算法分为四个步骤:首先对图像进行平滑滤波;之后计算水平方向和垂直方向的梯度幅值图像和梯度方向图像;然后根据梯度方向图像对梯度幅值图像进行噪声边缘滤波,并对梯度幅值图像进行非最大值抑制;最后对滤波后的梯度幅值图像进行分割,对得到的边缘映射图进行边缘跟踪,获取所有的边缘段及其特征,并滤除不满足特征约束的边缘,得到感兴趣的边缘。试验结果表明本文算法有很好的适用性。It is difficult to select the threshold and sometimes it introduces many disturbance edges in edge detection. Aiming at these problems, an interesting edge detection algorithm is proposed. The algorithm has four steps. Firstly, the image is smoothed to de-noise. Then, gradient magnitude image and gradient direction image are calculated. After that, gradient magnitude image is filtered according to gradient direction image. At last, gradient magnitude image is segmented to get initial edges ,and the edge mapping images are traced. Experiment results indicate that this algorithm has good applicability.

关 键 词：多级滤波 边缘检测 梯度幅值 梯度方向 边缘跟踪 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]