一种基于颜色和几何特征的人脸检测方法

A Method to Detect Human Faces Using Color and Geometrical Information

该文提出了一种有效的人脸定位方法。通过局部自适应门限提取图像边缘,用几何规则和颜色特征筛选出候选眼睛块,使用遗传算法将这些眼睛块配对验证,适应度函数采用一个新颖的眼睛及脸颊区域的灰度投影及肤色信息定义。该方法可以实现多尺度、多角度的正面人脸检测,实验结果令人满意。

In this paper,a new method is proposed to detect human faces in color image.In preprocessing stage,the edge map can be achieved by LAT,and then some rules with the geometry and color information are used to select the candidate eye pairs.In genetic algorithms ,these eye pairs are encoded.The fitness function is constructed according to the intensity and color information in eyes and cheek region.Experimental results show that the proposed method is efficient ,which can detect the frontal human faces with various orientations.