基于局部阈值的Canny边缘检测算法

Canny Edge Detection Algorithm Based on Local Threshold

传统Canny边缘检测算法采用全局高低阈值,只能检测全局显著而不能检测局部显著的边缘。另外,传统Canny算法需要人为设置阈值,使其难以适应不同图像的边缘检测。本文提出了一种改进的Canny算法,首先采用最大类间方差法获取梯度幅值图的自适应全局分割阈值,计算大于该阈值的像素点数与图像总像素点数的比值作为边缘点的全局比例值;其次在图像上每隔相同间隔选择1个像素点作为计算点,根据全局比例值计算获得每个计算点的局部高低阈;最后通过避免重复计算和插值计算方式加速局部阈值矩阵的计算。实验结果显示,本文Canny算法不仅能够自适应检测出图像中的局部显著边缘,而且具有较好的加速计算性能。

Traditional Canny edge detection algorithm can detect edges of global saliency rather edges of local saliency through global threshold.In addition,the need of setting the threshold artificially makes it hardly suitable for edge detection of different images.Hence,an improved Canny algorithm is proposed in this paper.Firstly,adaptive global segmentation threshold of the gradient magnitude image is obtained by Otsu method and then the ratio of the number of pixels larger than the threshold to the total number of image pixels is calculated as global proportion value.Secondly,pixel points are selected as computational points at the same interval of the image.Local high and low thresholds of each calculation point are calculated in accordance with global proportion value.Finally,the calculation of local threshold matrix is accelerated by avoiding repetitive computation and adopting interpolation.The experimental results show that Canny algorithm proposed in this paper can not only adaptively detect the local salient edges in the image but also achieve better performance of accelerating computation.