基于非单色背景下的人脸彩色图定位算法与实现技术

Algorithm and technology of localiza tion of color image of face based on multi-color background

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阚值,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限坐标,将人脸图象规范化。实验结果证明了该算法的有效性。

In this paper, we introduce a new Algorithm of localization of face, in the case of multi-color background face. This algorithm includes two steps. Firstly, iterative algorithm is used to compute the cb, cr threshold and automatically segment image. In this way, face image can be totally segmented. Secondly, the face is localized. Because the the cr threshold of lips is high, we can exclude the affect of the lips by delete the lips part. We can also fix the eye coordinate by using horizon and vertical gradation feature. The experiment proves the feasibility of the algorithm.