摘要
通过分析归一化相关匹配算法的原理,提出了建立累加和表与平方求和表来减少匹配过程中加减法的冗余计算量,并采用快速傅里叶变换算法减少乘法的计算时间,使算法在无损匹配性能的条件下大大降低了计算时间。同时,该快速算法对于匹配模板和搜索区域的增加不敏感,十分有利于工程实现。
Normalized Cross Correlation （NCC） algorithm is robust, but it has poor real-time performance due to its large amount of computation. By analyzing the principle of the algorithm, we present a fast NCC algorithm, which reduces the redundancy computation of addition and subtraction by designing accumulation sum table and square sum table. At the same time, FFT algorithm is adopted to decrease the computational cost of multiplication. In addition, this fast NCC algorithm is lossless to the capability of correlation, and insensitive to the increment of template size or search region, so that it is quite fit to practice in reality.
出处
《信息与电子工程》
2006年第6期461-463,共3页
information and electronic engineering
关键词
图像匹配
快速傅里叶变换
实时
归一化
image matching
Fast Fourier Transform（FFT）
real time
normalized
作者简介
陈松柏（1965-），男，四川省雅安市人，四川农业大学讲师，研究方向为电子技术应用.Email：csb719@sohu．com，
