为克服传统相关匹配算法计算复杂度过大的缺点,基于快速傅里叶变换和积分图提出了一种快速相关匹配算法.在不改变传统相关匹配计算结果的前提下,该算法通过对相关系数公式的分析和化简,用快速傅里叶变换计算模板图像和基准图像的相关,并采用积分图计算基准图像灰度值及其能量的窗口积分,使相关匹配计算量与模板图像大小近似无关,从而大大降低了计算复杂度.不同光照条件下图像匹配实验和复杂环境下目标跟踪实验结果表明,该算法具有很强的抗干扰能力,其执行效率比传统相关匹配提高1~2个数量级.

In order to overcome the computation complexity of traditional correlation matching algorithm, a fast correlation matching algorithm is proposed based on fast Fourier transform and integral image. By analyzing and simplifying the equation of correlation coefficient, fast Fourier transform is employed to calculate the correlation between the template image and the base image, and integral image is employed to compute the window integral of the gray value of the base image and its energy. The computation complexity of the proposed algorithm is approximately independent on the size of the template image and thus is greatly decreased, without changing the results of traditional correlation matching. Experimental results for image matching under different illumination conditions and object tracking in complex situations show that the proposed algorithm is robust to interferences, and its implementation efficiency is 1 to 2 order times faster than traditional correlation matching algorithm.