适于人脸检测的模板匹配快速算法

为提高人脸检测中模板匹配的速度,在积分图像基础上,提出和应用平方积分图像,实现了一种计算速度与图像窗口大小无关的快速灰度均值和方差算法。应用这种快速算法,结合对灰度分布标准化近似计算公式、相关系数和平均偏差计算公式的变换,以及灰度分布标准化处理时省略拷贝图像窗口及窗口灰度值变换等不必要的步骤,大大提高了模板匹配速度。人脸检测实验证明了这种算法的有效性。

Rapid algorithm of template matching for face detection

In order to improve the speed of template matching in face detection, based on integral image, a square integral image was proposed and used to realize a rapid algorithm for calculating grayscale average and variance of image windows, and the calculating speed was independent on the windows size. With this rapid algorithm and transformation of some formulas, such as approximate formulas for computing grayscale distribution normalization, formulas for computing the correlation coefficient and average deviation, and leaving out some needless processes such as copying windows and conversion of windows grayscale value, the template matching was sped up greatly. Experiment results of face detection demonstrate the efficiency of this algorithm.