改进SSR算法在人脸识别中的应用

Application of Face Identification Based on the Improved Algorithm of SSR

为提高人脸检测中人脸验证的速度,在平方积分图像和图像窗口的灰度均值及方差快速计算基础上,对传统的SSR算法中人脸验证算法部分进行了改进。应用改进的算法,结合窗口灰度标准化的计算和平均人脸模板的取得,引入相关系数和平均偏差,加快了求解速度,克服了由于噪声点的存在对真实的候选脸图像漏检的情况。人脸检测实验证明应用改进的模板匹配算法的正确性.

In order to improve the speed of face verification in face detection,based on a square integral image and a rapid calculating for grayscale average and variance of image window,traditional algorithm of SSR in face verification was improved. With this improved algorithm,combined with calculation of window grayscale normalization and acquisition of the average face template,the introduction of the correlation coefficient and average deviation,the solving was sped up greatly and the neglected check for real candidate face image was avoided,even in existence of noise. Experiment results of face detection demonstrate the correctness of the template matching algorithm.