一种基于变点统计分析的图像边缘检测的方法

Edge detection method based on nonparametric change point theory

基于统计的图像边缘检测方法是计算机视觉中边缘检测的重要方法之一。提出了一种基于非参数变点统计分析的方向性边缘检测算法,该方法可以最小化对图像数字特征的先验信息的需求。深入讨论该算法在含有高斯噪声和椒盐噪声的灰度图像处理中的一些问题,通过实验与MATLAB的经典的边缘算子sobel算子和canny算子相比较。该方法不仅能很好地检测出图像的真实边缘,而且有效地抑制了两种噪声对边缘检测的影响,取得了较好的效果。

Statistical method is one of important methods on edge detection of computer vision.This paper proposes a method of edge detecting which is based on the change point statistic analysis because it minimises the need for a priori information about images.What＇s more,some problems for gray level images is studied to which Gaussian white noise and salt and pepper noise are added in this method.Compared with well-known sobel algorithm and canny algorithm edge detectors in MATLAB,this approach not only detects real boundary of image,but also effectively suppresses the impact of two types of noises on edge detection.This new approach makes a great progress.