一种基于最小交叉熵的Canny边缘检测算法

摘要

The traditional Canny edge detection algorithm exists the defect losing details of the edge, so this paper proposed an improved Canny edge detection algorithm based on minimum cross entropy. High and low threshold are given an fixed value in traditional Canny operator, that may cause much false edge information when the image's all gray level concentrates on one region. It may get a desired high threshold value based on minimum cross entropy , and the Gaussian filter parameter can be calculated according to the gray's mean value and variance. The experiment results indicate that the algorithm improves the effect of edge detection in the case of noise disturbance, and has a good robustness.%考虑传统Canny算子在边缘检测中的不足,提出一种融入最小交叉熵的Canny边缘检测算法.传统Canny算子的高低阈值一般人为地设定固定值,当一幅图像的灰度级集中在某一区域时,容易造成虚假边缘；利用最小交叉熵计算图像的高低阚值,可以得到理想的阈值；利用图像灰度的均值和方差计算Canny算子的高斯滤波参数.实验结果表明,算法较好地提取了图像的边缘信息,抑制噪声能力较强,有效地提高了边缘检测的鲁棒性.