快速图像匹配相关系数算法及实现  

A Fast Algorithm of the Correlation Factor in Image Matching and Implementation

作　　者：刘红侠[1] 杨靓[1] 黄巾[1] 黄士坦[1] 

机构地区：[1]西安微电子技术研究所,陕西西安710054

出　　处：《微电子学与计算机》2007年第2期32-35,共4页Microelectronics & Computer

基　　金：国家"十五"预研课题(413160203)

摘　　要：最大归一互相关图像匹配算法是图像匹配中的常用算法,其关键是解算活动图与基准图间的相关系数。针对相关系数计算量大的特点,分析了FFT的基与FFT处理速度之间的关系以及基16FFT算法特点,提出用基16FFT算法计算相关系数,相关系数的处理时间大幅减小;同时针对高基蝶形单元设计复杂、使用不灵活等特点,提出采用级连思想实现主基16蝶形单元,使处理器的设计复杂度降低。实验证明,将主基16FFT处理器用于相关系数的计算中,使最大归一互相关图像匹配处理速度达到国际领先水平。The maximum correlation algorithm （MAC） is a traditional algorithm in image matching. It＇s key lie in calculating the correlation factor between reference image and real-time image. To the huge amounts of calculation of correlation factor, the relationship between radix and the speed of FFl＂ algorithm and the feature of radix-16 FFl＂ algorithm are analyzed in this paper. As a result, a new method of computing the correlation factor with mainly radix-16 FFT algorithm is proposed. This method makes the number of calculation decreased and the processing of correlation factor speeded up and the design complication of radix-16 butterflier decreased. It＇s validated that image matching speed can be increased by using a mainly radix-16 FFT processor to compute the correlation factor in image matching system.

关 键 词：最大归一互相关算法 主基16FFT算法 相关系数计算 主基16FFT处理器 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论] TP338.7[自动化与计算机技术—计算机科学与技术]