基于Canny算子与阈值分割的边缘检测算法

The edge detection algorithm based on Canny operator and threshold segmentation

针对传统Canny算子阈值选择困难的问题,提出一种基于最小均方误差计算高低阈值的方法.首先对采集到的图像进行增强处理,提高对比度,然后把图像中的灰度作为模式特征.假定各个模式中灰度的随机变量是独立分布的,根据增强后图像的目标和背景模式服从高斯分布的特征,通过两个概率密度函数的解析式,从而得到Canny算子的高、低阈值.最后,将边缘检测算法与Canny算子提取边缘算法、最小均方误差法提取边缘算法进行相互比较,结果表明,采用本文的算法提取图像的边缘更加清晰有效.

In view of the threshold selection difficulty of traditional Canny operator,a method to calculate high and low threshold based on minimum mean square error was proposed.First of all,in order to improve the contrast ratio,the collected images was enhanced.And then the gray degree as characterized model was selected,assuming the random variable of each pattern of gray was independent distribution.Since the enhanced images of target and backgrounds obeyed the Gaussian distribution characteristic,high and low threshold could be received by two analytic formula of probability density function.Finally,compared with the method of Canny operator and minimum mean square error,the results showed that the proposed method could extract the edge of information more clearly and effectively.