高斯-拉普拉斯边缘检测算子的扩展研究

Extended Laplacian of Gaussian operator for edge detection

针对经典的高斯-拉普拉斯(LOG)边缘检测算子是各向同性的,对各个角度方向的图像边缘检测的力度是相同的特性,对经典LOG边缘检测算子引入了角度信息参量进行推导,使以圆为对称的经典的LOG边缘检测算子变成为以椭圆对称,并且可以在坐标轴旋转任意角度的边缘检测算子,增强了其边缘检测的功能,使之能对不同角度方向的边缘更加有效地进行检测.经过在Matlab里对同一幅图像进行比较实验,对于图像中不同角度的边缘均能相应地进行提取.扩展后的LOG算子,不仅增强了边缘检测算法功能,而且完全保留了经典LOG算子原有的优点.

Being an isotopic, classic Laplacian of Gaussian（LOG） edge detect operator only can detect image edges in all directions isotopically. The parameters of angle information were introduced into classic LOG operator for edge detection and deduced mathematically. It is symmetrical as ellipse instead of circle, and can detect models in rotary angle edge by the coordinate axis. Moreover, it can effectively detect the edges in different orientations that strengthen the edge detection capability. After comparative experiment on the same image in Matlab, it presents that edges in different angles can be achieved. The extended LOG operators not only strengths the edge detection function, but also preserves all the advantages of the former classic LOG.