基于多方向的各向异性边缘检测算法

针对传统边缘检测算法抗噪性差的不足,提出了一种多方向的各向异性边缘检测算法。该算法构造了4个具有各向异性的5阶差分模板,对其进行归一化处理后,分别对待检测图像进行卷积处理,根据检测算法在各方向上卷积结果的幅值和方向信息得到灰度边缘图,最后采用最大类间方差法确定阈值进行边缘二值化。多组仿真实验结果表明,该方法能有效实现边缘提取,比传统方法具有更高的检测精度和更强的噪声鲁棒性。

Anisotropic Edge Detection Algorithm Based on Multi-Direction

The traditional edge detection operators have sensitivities to noise,an anisotropic edge detection algorithm based on multi-direction is proposed in this paper.Firstly,four anisotropic fifth-order differential templates are constructed,then the pending image with convolution after being normalized,the gray edge image will be obtained by magnitude and direction information of convolution results based on detection operator in all directions,and finally utilizes the method of maximum classes square errors to determine a threshold and extract binarization edges.Several sets of simulation results demonstrate the feasibility and effective ness of the proposed algorithm,it can obtain clear edge and high detection accuracy,m eanwhile,the algorithm has better ability of preserving the edge details and better robustness to noises than traditional methods.