一类细节可控的边缘检测算子

A CLASS OF CONTROLLABLE EDGE\|DETECTING OPERATOR

Ｍａｒｒ图像边缘检测的基本思想是先利用高斯平滑算子对原始图像进行平滑，然后对平滑图像求拉普拉斯变换的零交叉；但是利用高斯平滑算子平滑的图像和原始图像存在着误差，而误差图像残留着一些边缘信息，而这些残余信息可能造成检测到的边缘线不连续和有用的特征点可能检测不到．针对这个问题，文中根据Ｍａｒｒ边缘检测基本思想，导出了一类细节可控的边缘检测算子，其中Ｍａｒｒ算子只是它的一个特例．理论和实际数值实验结果表明，这类边缘检测算子有一定的实际效果和参考价值．

Based on Marr's image edge detecting theory and the thought that the residual between the original image and smoothed image remain some feature point, a class of controllable edge detecting operator is constructed in this paper. Practical numerical experimental results show that these edge detecting operators have some practical effect and reference value.