Face normalized processing is the key step of face recognition pretreatment, so this paper presents a face normalization algorithm based on the localization of eyes and mouth. First, eliminate the influence of uneven illumination through histogram equalization, and then find the localization of eyes and the center of mouth through gray projection method. Then according to the localization results of eyes, do face in-plane rotation to realize geometrical position normalization by affine transformation. Finally, scale on the horizontal and vertical to realize size normalization in terms of the localization information of the eyes and the center of mouth. The computational complexity of the algorithm is small, simple and feasible, and it has good localization performance and the normalized effect.%人脸归一化处理是人脸识别的关键预处理步骤，本文提出了一种基于眼和嘴定位的人脸归一化算法。首先通过直方图均衡化法消除不均匀光照对人脸图像的影响；再通过灰度投影法对人脸的双眼和嘴部中心进行定位；然后根据双眼的定位结果，通过仿射变换法进行人脸平面内旋转，实现几何位置的归一化；最后根据双眼和嘴部中心的精确定位信息进行横向和纵向的缩放，实现大小的归一化。该算法运算量小、简单可行，并具有良好的定位效果和归一化效果。
展开▼
