肤色与梯度互补的人脸检测研究  

RESEARCH ON FACE DETECTION BASED ON COMPLEXION AND GRADIENT COMPLEMENTARY

作　　者：李岚[1] 朱林杰[2] 张云[1] 

机构地区：[1]甘肃联合大学电子信息工程学院,甘肃兰州730000 [2]河南理工大学后勤集团,河南焦作454000

出　　处：《计算机应用与软件》2013年第2期280-282,295,共4页Computer Applications and Software

摘　　要：用YcbCr色彩空间中的Cr、Cb通道和HSV色彩空间中的H通道相结合,确定一种混合的肤色模型,即三维直方图模型。肤色在三维直方图中的聚类进行投影,结果为一个椭球,从而检测出可能的人脸区域。根据人脸的边缘形状,建立梯度模型定位人脸位置。利用两种模型的互补性,对算法进行线性组合,确定一个匹配度函数f(k)来衡量是否检测到目标人脸。实验结果表明该方法能够很好地排除人体其他部分裸露的肤色和背景中类肤色的干扰。其运算简单,检测速度快,对于侧面人脸和部分遮挡也有很好的效果。A mixed complexion model,the three-dimensional histogram model,is determined by combining the Cr,Cb channel of YcbCr colour space and the H channel of HSV colour space.The projection of the cluster of complexion in three-dimensional histogram forms an ellipsoid,and is used to detect the possible face regions.According to the edge shape of the face,the gradient model is established to locate the face position.The matching degree function f（k） is determined to measure the target face using the linear combination of the algorithm based on the complementary of these two models.Experimental results show that this method can exclude the interference of other exposed body parts and the background colour.It has the simple operation,quick detection rate and good results to the side face and partial occlusion.

关 键 词：肤色采样 混合肤色模型 梯度模型 人脸检测 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]