基于非单色背景下的人脸彩色图象定位算法与实现技术

Algorithm and realization human face location in non-single color backgrounds

针对非单色背景下的人脸,本文提出了一种新的人脸定位算法。该算法包括两个部分:首先进行肤色分割,该文采用迭代算法计算Cb、Cr阈值,进行自动分割,可完整地分割出人脸部分;然后进行人脸定位,由于嘴唇具有较高的Cr值,可去除嘴唇部分以消除嘴唇对其它图象部分的影响,利用水平、竖直灰度特性确定人眼坐标,将人脸图象规范化。实验结果证明了该算法的有效性。

A new face location algorithm is put forward to deal with Non-single Color Background. The algorithm includes two parts: Firstly, a color segmentation algorithm is put forward, iterated algorithm is applied to calculate Cb and Cr threshold values of the shin color, segmentation is completed automatically. Secondly, a face location algorithm is put forward.As the lips have comparatively high Cr threshold value, the algorithm gets rid of the part of lips,in order to eliminates the influence of lips on the other parts of the picture, it uses the properties of horizontal and vertical gray degrees to fix the coordinate of the eyes, then, it makes the face picture standard. Experiment results confirm the efficiency of algorithm.