基于相位一致性原理的特征点检测  

Novel feature points detection based on phase congruency principle

作　　者：王怀所 黄华[1] 

机构地区：[1]四川大学电气信息学院,四川成都610065

出　　处：《计算机工程与设计》2014年第7期2427-2433,共7页Computer Engineering and Design

摘　　要：为了从原始图像中提取具有较强表示能力的图像特征,研究了相位一致性原理,提出了基于相位一致性原理的特征点检测算法。在相位一致性原理的基础上,代替DoG和Harris特征点检测算子,提取了准确稳定的图像特征点。在介绍了DoG和Harris相关算法原理后,将基于相位一致性原理的图像特征点同DoG、Harris提取的特征点进行了主观和客观的对比分析,显示出相位一致性原理特征点具有特征点位置准确、分布广泛、可重复性高、算法稳定等优势。结合实例验证了该算法的有效性和可行性。To extract good image feature points with strong ability to represent an image from an original image,the phase con-gruency principle was studied deeply and the algorithm based on the phase congruency principle was presented.The algorithm was based on the phase congruency principle instead of DoG and Harris feature point detection operator to extract the accurate and stable image feature points.After introducing DoG and Harris algorithm principle,the comparisons of the feature point ex-traction among the phase congruency principle,DoG and Harris were analyzed subjectively and objectively.The experimental re-sults showed that the phase congruency principle based feature point detection was feature-points-position accurate,widely dis-tributed,highly repeatable and stable.Finally,the performances and the feasibilities of the proposed method were demonstrated by the example.

关 键 词：相位一致性 关键点 高斯差分 HARRIS算子 特征点检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]