基于八方向卷积模板的边缘检测算法

Edge detection algorithm based six directions convolution templates

针对传统图像边缘检测算法存在的边缘方向性不强及边缘较粗等问题,提出了一种基于八方向卷积模板的边缘检测算法。算法采用0°、22.5°、45°、67.5°、90°、112.5°、135°和157.5°八个方向的卷积模板进行边缘检测,模板权值根据中心像素点到邻域像素的距离及方向夹角的大小进行设定,充分考虑到了邻域内像素对中心点方向梯度的贡献大小,能够较好地检测出图像不同的方向边缘。对梯度图像采用了改进的非极大值抑制方法进行细化,可得到单像素的图像边缘。实验结果表明,该算法获取的边缘图像边缘较为完整,方向性强且边缘较细,整体效果明显优于传统Sobel算法。

An edge detection algorithm based on six directions convolution templates is proposed aiming at the shortcoming of traditional Sobel operator in poor location accuracy and coarse edge. The algorithm uses six direction templates of 0° ,22.5° , 45° ,67.5° ,90° ,112.5° ,135° and 157.5° to detect image edge, considering the contribution of neighborhood pixels for the center pixel in the direction gradient, the templates weights is determined according to the distances and the direction included angle of neighborhood pixel and the center pixel, the different image edge can be detected well. In order to get single pixel width edge, uses improved non-maximum suppression to refine the edge of gradient image. The results show that the new algorithm can detect the image edge relatively complete,with accuracy direction and thinner edge, the whole effects are better than traditional Sobel algorithm.