一种基于统计的图像边缘检测方法

Statistical approach of edge detection

提出了一种基于统计的图像边缘检测方法。它首先通过计算图像中每一点的梯度向量以及方差—协方差矩阵,然后对每一点的梯度向量进行非极大值抑制,最后使用统计的方法求出局部标准化梯度阈值来检测图像的真实边缘。通过与经典的边缘算子相比较,该方法不仅能很好地检测出图像的真实边缘,而且有效地抑制了虚假边缘的产生,取得了良好的效果。

A statistical approach of edge detection was presented. First, this approach determined gradient vector and variancecovariance. Second,it made use of non-maxima suppression to every gradient vector. Last, local standardization for gradient threshold base on statistical approach was computed and use it for edge detection. To compare with well-known edge detectors, this approach not only detects real boundary of image, but also effectively suppresses the false edge of image. This new approach receives a good effect.