基于梯度方向的零交叉边缘检测

Zero Crossing Edge Detection Based on Gradient Directions

摘要:

鉴于高斯拉普拉斯(LoG)算子具有各向同性且在有方向性差异的场合并不适应的不足, 提出在梯度方向进行零交叉的边缘检测方法。首先使用自适应平滑滤波原图像; 然后在水平和垂直两个方向上分别计算图像的一阶梯度分量; 使用梯度算子分别对两个求得的一阶梯度分量进行操作, 计算两方向上的二阶偏导数; 分别对两个方向上的二阶偏导数进行零交叉检测, 并合并零交叉点得到边缘图像; 最后使用形态学图像处理方法去除边缘图像中面积较小的孤立点区域。结果表明, 该方法在噪声环境中具有良好的边缘检测效果, 且运算时间与LoG算子检测方法相当。

Abstract:

Owning to its isotropy, traditional LoG operator can not be used in many occasions with anisotropy. So the method of zero crossing detection based on gradient directions for edge detection was presented. Firstly, adaptive smooth filter was used to smooth the images, and then the gradient components were calculated in both horizontal and vertical directions, respectively. And then, the second order partial derivative of the image was calculated using the first-order gradient operators, after which, zero crossing detection was performed on the edge points in the two directions and all the detected edge points were combined to obtain edge images. Lastly, mathematical morphology method was used to remove isolated regions with small area. The experimental results show that the method has good effect of edge detection in the noise environment, and also its computation time is almost the same as that of LoG operator.