基于Sobel的多方向算子模板边缘检测算法

Sobel-based edge detection algorithm for multi-direction template

针对传统Sobel算子存在的边缘检测方向性不强及提取边缘较粗等问题,提出了一种改进的多方向算子模板的边缘检测算法。算法增加了22.5°,45°,67.5°,112.5°,135°和157.5°六个方向算子模板,能够较好地检测出图像不同的方向边缘。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;算法对梯度图像采用了改进的非极大值抑制方法进行细化,得到了较细的图像边缘。实验结果表明,与传统Sobel算法相比,该算法提取的边缘图像具有边缘方向性强且边缘较细的优点,具有较高的应用价值。

Since the traditional Sobel operator exists the weak edge detection directivity and the extracted edge is coarse,an improved edge detection algorithm of multi-direction operator template is proposed. In the algorithm,six direction templatesare added at 22.5°,45°,67.5°,112.5°,135° and 157.5° to detect different direction edges. The templates weights are deter-mined according to the distances from center pixel to neighborhood pixels and their direction included angles,which fully considerthe contribution of neighborhood pixels to the central point direction gradient. The improved non-maximum suppression methodis used in the algorithm to refine the gradient image for getting the thinner edges. The results show that the improved algorithmhas advantages of stronger edge directivity and thinner edges in comparison with traditional Sobel algorithm. It has a high appli-cation value.