摘要
在Hsu,Abdel-Mottaleb and Jain提出的检测算法(简记为Hsu法)的基础上,提出一种改进的适用于彩色图像的人脸检测算法。使用相关技术对原图像进行预处理,采用高斯分布模型代替椭圆参数模型得到原始图像的肤色概率图,在不影响准确度的前提下提高了检测速度;采用自适应阈值和直方图法相结合选取肤色分割阈值更准确;根据人眼信息在子图像上进行人眼检测。经实验验证,该算法能较好地检测出复杂条件下的人脸,与Hsu法相比速度有较大的提高。
Based on the related technique proposed by Rein-Lien Hsu, Mohamed Abdel-Mottaleb and Anil K. Jain ( denoted as Hsu's method simply), an improved algorithm of face detection was proposed for color images. The original images were preprocessed by the related technique. In order to reduce the computational complexity of Hsu's method, the parametric ellipse mode was replaced by the Gaussian distribution model, which led to the fast face detection without affecting the detection accuracy. A method of selecting the skin threshold based on the combination of self-adaptive process and histogram was more accurate. In order to extract the face areas accurately, the eyes were detected in the previously extracted subimages. Experimental results demonstrate successful face detection under complex conditions. The proposed method is much faster than Hsu's method.
出处
《计算机应用》
CSCD
北大核心
2008年第4期986-989,共4页
journal of Computer Applications
基金
国家自然科学基金资助项目(60472060
60572034)
2006年教育部新世纪优秀人才计划项目
江苏省自然科学基金资助项目(BK2006081)
关键词
人脸检测
高斯分布模型
人眼检测
face detection
Gaussian distribution model
eye detection