摘要
针对人脸检测中的非线性和降维问题提出了一种有效的人脸检测方法。首先基于人脸的肤色和几何特征,通过局部自适应门限方法提取图象边缘,用几何规则和颜色特征筛选出候选眼睛块。然后采用改进的四边界sobel算子对眼睛精确定位。最后使用训练好的支撑向量机来验证。该方法可以实现多尺度多角度的彩色图像中的正面人脸检测,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 geometrical and color information are used to select
the candidate eye pairs. Then the orientation of eyes could be got with a novel sobel operator. At last the trained
support-vector machine is used to certify the candidate eyes blocks. Experimental results show that the proposed method
is efficient, which can detect the frontal human faces with various orientations.
出处
《内江师范学院学报》
2004年第2期14-18,共5页
Journal of Neijiang Normal University
关键词
肤色检测
支撑向量机
人脸检测
灰度投影
support vector machine
face detection
intensity projection
edge detection
