基于肤色分割与几何特征的人脸检测

Face detection based on skin color segmentation and geometry information

针对传统的人脸检测方法存在算法复杂、鲁棒性差以及精确度低等不足,提出一种基于肤色分割与几何特征的人脸检测算法。首先,将人脸彩色图像进行光照补偿预处理与非线性转换,建立YCb Cr颜色空间肤色模型同时融合人脸几何特征进行粗定位;然后,在人脸区域内,依据眼睛与嘴巴的色度特性分割出眼睛与嘴巴的特征区域并确定各自特征点的中心位置;最后利用眼睛,嘴巴的几何关系构建三角形精确定位人脸。实验结果表明,该算法简单,在速度与准确性方面具有良好的性能。

In view of the high computational complexity,poor robustness and low accuracy of traditional algorithm for face detection,this paper proposes a novel algorithm for face detection based on skin color segmentation and geometric feature. Firstly,YCb Cr color space model is established and it combines geometrical characters to achieve the purpose of coarse detection with lighting compensation technique and non-linear transformation to face color image. And then owing to chromaticity property of eyes and mouth in the face feature region to implement the segmentation of eyes and mouth and it determines the center position of each feature points. Finally,it locates the face precisely according to the geometrical relationship of eyes and mouth to form triangle. The experiment demonstrates that the proposed method is not only simple,but good performance in detection speed and accuracy.