基于积分投影与灰度复杂度的人脸特征定位算法

Human Face Characteristic Location Algorithm Which Based on Integral Projection and Grey Scale Complexity

针对当前人脸特征定位的研究现状与难题,采用YCbCr椭圆聚类方法进行肤色区域提取,然后根据人脸的几何特征,通过比例、大小结构特性来判断二值化图中的目标区域是否为人脸,从而可以排除非人脸区域,再结合人脸面部器官分布特点,利用积分投影方法,对人脸区域垂直和水平两个方向分别进行灰度值累加统计,经过分析可以定位眼嘴的区域边界和中心点。同时针对积分投影的缺点,即如果头部发生倾斜会导致投影位置不准确,提出了基于灰度复杂度的眼睛定位方法,利用该方法首先定位人眼位置,然后根据人眼对人脸区域旋转校正,最后再利用积分投影定位嘴巴。

Being aimed at the study situation and problems of human faces characteristic location, the paper adopt the means of YCbCr clustering to extract from the zones of face color. And then according to the geometric characteristic, it can analyze the pro- portion and size structure in the goal area of binarization pictures, excluding impossible the zones of human face. Through the feature of human-face-organ distribution, further more ,it can take advantage of the means of integral projection to add up the two directions -- vertical and horizontal on human faces zones .It can position edge of area and center point between eyes and mouth through analyzing. At the same time, because of the utmost disadvantage of integral projection, namely, projection location is not accurate if the head slope away, the means of eyes location which is based on grey scale and complexity is raised. Take advantage of this means, it location the eyes firstly, and then arrording to the eyes location to rotate and correct human-face-zone and location the mouth.