一种图像亚像素边缘检测算法的改进研究

Research of an improved algorithm about sub-pixel edge detection for images

为了提高刀具预调测量仪的检测精度,提出了一种改进的图像快速亚像素边缘检测算法——基于正交多项式拟合的亚像素边缘检测算法。首先,利用传统的Sobel算子完成边缘点整像素级别的检测,确定边缘的主体区域;然后,过边缘点沿边缘法线方向拓展像素,取一系列像素点并计算其灰度值;最后根据像素点灰度分布的数学特征,利用正交多项式和最小二乘法求拟合函数,通过拟合曲线确定图像边缘点的精确位置,实现图像亚像素边缘检测。实验证明,该算法运行时间短,约为0.63 s;检测精度高,可达0.1 pixels。

In order to improve the detection accuracy of tool presetting and measuring machine, an improved sub-pixel edge detection algorithm for images is proposed. Firstly, the Sobel algorithm is carried out to get some edge points at pixel level, and then find the approximate position of edge. Secondly, along the normal of the edge direction with edge points, the pixel is expanded and its gray is calculated. According to mathematical feature of gray distribution of points, the gray level curve is fitted through orthogonal polynomial and least square fit method and the accurate position of edge point with a sub-pixel precision is achieved. The simulated experiments results of MATLAB show that the proposed novel algorithm needs much shorter running time, about 0.63 s, and much higher precision, about 0.1 pixels.