梯度图像的阈值选取

Threshold Selection of Gradient Image

为了较精确地提取目标的边缘轮廓,针对梯度图像的特征,提出在canny边缘检测算法基础上,采用自适应阈值消除噪声,探测较精确的边缘轮廓。由于光学成像和高斯平滑滤波等作用,梯度图像中边缘上噪声的分布是呈近似单边正态分布,所以根据梯度图像的灰度均值与方差确定自适应阈值。通过实验仿真比较,采用自适应阈值获得了较准确的边缘轮廓。

In order to extract edge contour of the object precisely,based on the characteristics of gradient image,an adaptive threshold was put forward to eliminate noise and detect more accurate contour on the basis of canny edge detection algorithm.Because of the optical imaging and adge Gaussian average filtering,the distribution of noises on the edge of the gradient image was approximate to the unilateral normal distribution.Adaptive threshold was determined according to gray mean and standard deviation of the gradient image.The simulations proved that more accurate edge contour is detected with the adaptive threshold.