一种基于区域分割的无阈值Harris特征点检测算法  

A Region-based Segmentation no Threshold Harris Feature Point Detection Algorithm

作　　者：张麟华[1] 刘鹏[2] 杭鹏越 岳晓冬[3] 

机构地区：[1]太原工业学院计算机工程系,山西太原030008 [2]中国电子科技集团公司第二研究所,山西太原030024 [3]上海大学计算机工程与科学学院,上海200444

出　　处：《山西大学学报（自然科学版）》2014年第1期57-63,共7页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(61103067)

摘　　要：文章针对传统Harris算法需人为设定阈值和特征点聚簇的问题,实现了一种基于区域分割的无阈值Harris特征点检测算法,在图像区域分割过程中引入了极差以剔除不存在特征点的区域,采用了计算区域信息熵差值的方式以减少噪点的干扰,针对10×10像素分块后特征点仍然过密的情况采取了四块合一块的方法,最后根据实际情况在临近的特征点中只保留特征值最高的。实验结果表明,与传统的Harris特征点检测算法相比,本文算法避免了手动设置阈值的不确定性,而且特征点分布均匀、合理,没有出现特征点聚簇的现象,同时具有一定的抗噪性。Traditional harris algorithm needs artificial threshold and exists the problem of clustering feature points, to solve these issues,the paper implements a region-based segmentation no threshold Harris feature point detection algorithm. The range is introduced in the region segmentation to exclude the area of the fea- ture points which are not exist. Regional entropy gradient calculation methods is adapted to reduce noise in- terference. Taking four blocks together is described to solve the problem of feature points denseness after the block of 10 × 10 pixels. Finally,according to the actual situation,only the feature point which has the highest value in feature points neighborhood is retained. Compared to the traditional harris feature point detection algorithm,the experimental results show that the algorithm is not only avoiding the uncertainty of a manually set threshold,but also causing feature points distribution,reasonable,and no clustering phe- nomenon,and having a better noise immunity.

关 键 词：特征点检测 区域分割 HARRIS算法 熵值差 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]