一种基于统计特征的孤立点和边缘点检测算法  

An algorithm of detecting outliers and edge points based on statistical distribution feature

作　　者：牛立尚[1] 

机构地区：[1]辽宁抚顺职业技术学院基础部,辽宁抚顺113122

出　　处：《信息技术》2015年第6期112-114,共3页Information Technology

摘　　要：提出一种基于统计分布特征的图像孤立点和边缘点检测的方法。通过定义离群半径,将每个像素点的邻域点进行初步分类,得到疑似孤立点,然后根据每一个点在不同邻域被划分为疑似孤立点的次数,计算该点的孤立置信度。根据孤立点和边缘点具有不同的邻域无关性的特性,从而可以用不同的孤立置信度阈值将两者分别检测出来。实验表明,文中的方法能够有效地检测孤立点和边缘点。An algorithm of detecting outliers and edge points based on statistical distribution feature was proposed. The neighborhoods of each point in the image are firstly classified based on a predefined outlier radius,and a set of suspected outliers are obtained. Then the outlier confidence of each point is calculated by counting the times which the point is regard as suspected outlier in different neighborhoods.Due to the different characteristics of neighborhood independency between outliers and edge points,the outliers and edge points can be separately detected by using different confidence threshold. The experiments show that the algorithm can detect the outliers and the edge points effectively.

关 键 词：孤立点检测 分布特征 疑似孤立点 离群半径 孤立置信度 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]