一种改进的Canny边缘检测算法

摘要

针对经典Canny边缘检测算法的一些问题,提出一种既具有较好的抗噪声能力又能自适应提取较多边缘细节改进的Canny边缘检测算法.该算法思想是先进行图像的预处理,包括基于模糊数学方法去除背景噪声的干扰,利用中值滤波和高斯平滑滤波去除图像中的随机噪声.然后,通过差分方法获得图像中各像素点的梯度矢量和梯度幅值,采用非极大值抑制方法获得极值点.最后,基于最大熵自动获取阈值,根据极值点图像和阈值得到边缘图像,使用边缘跟踪算法将不连续的边缘连接起来,并去除孤立噪声点.实验结果表明:改进算法所获得的边缘图像平均提高8dB,与传统方法相比,可获得更加连续的边缘和更少的噪声点.%We introduce an improved Canny edge detection algorithm for extracting much more edge information with less noise.Firstly,the background noises of images are eliminated using fuzzy mathematical method,and the random noises of images are removed using median filtering and Gaussian smoothing filter.And then,the gradient vector and gradient magnitude of every pixel point in images are obtained utilizing the difference method,and the extreme points are obtained using non-maximum suppression method.Finally,the edge images are achieved based on the extreme point images and thresholds which are automatically selected based on maximum entropy algorithm,and the edge images were further processed utilizing the edge tracing and noise eliminating methods.The results show that the signal to noise ratio of edge images obtained with improved algorithm are increased by 8 dB,compared with that of edge images obtained with Canny method.The improved Canny algorithm can obtain much more edge information and less noise than Canny algorithm.