一种自适应红外图像角点检测算法  

Adaptive corner point detection of infrared images

作　　者：邹志远[1] 安博文[1] 曹芳[1] 潘胜达[1] 

机构地区：[1]上海海事大学信息工程学院,上海201306

出　　处：《激光与红外》2015年第10期1272-1276,共5页Laser & Infrared

基　　金：国家自然科学基金(No.61171126);交通运输部应用基础研究项目(No.2014329810060);上海重点支撑项目(No.12250501500)资助

摘　　要：针对Harris角点检测算法自适应性差的问题,提出一种自适应角点检测算法。根据Harris算法定义的像素响应函数值的大小特性,得出阈值(Threshold)应满足的下限条件,继而对图像进行分块,得出每一块图像的阈值下限条件。综合考虑各图像块的阈值,采用加权方法得到图像总的阈值。采用局部保留最大响应值策略来避免角点聚簇的现象。试验结果表明:提出的自适应阈值计算方法在引入少量数学运算的前提下,使角点检测具有自动性,并且可以保证合理数量的角点,采用的剔除策略可以很好的避免角点聚簇现象,使图像最终角点数量合理、分布均匀。In order to solve the poor adaptability problem of the Harris comer detection algorithm, an adaptive comer detection algorithm is proposed. Firstly the condition which the threshold should meet the lowest value based on the definition of Harris comer is obtained. Then the image is divided into several parts and the threshold of every part is calculated according to the above condition. Finally the threshold of whole image is calculated by using weighting method after the threshold of every part is considered. Local retaining maximum response value method is adopted to avoid comer cluster. The resuhs show the proposed method can make comer detection has good automatieity through introducing a few arithmetical operation, and ensure the reasonable number of corner and avoid comer cluster.

关 键 词：红外图像 HARRIS角点 自适应 角点聚簇 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]