复杂背景中的人脸检测与定位  

Face Detection and Location in Complex Backgrounds

作　　者：孙怡[1] 吴艳冬[1] 胡家升[1] 

机构地区：[1]大连理工大学电信学院,大连116023

出　　处：《计算机工程与应用》2003年第1期88-90,98,共4页Computer Engineering and Applications

摘　　要：基于人类视觉机制,描述了一种由粗到细的复杂背景中的人脸检测方法。其基本思想是首先粗略定出人脸可能存在的区域,然后在可能的区域内进一步细致匹配,以证实人脸的存在并对其进行准确定位。在粗略检测中利用人脸的肤色分布统计模型将人脸从背景中分割出来;在准确定位中联合人脸的肤色分布统计模型、发色分布统计模型以及不同方向的头部模型,用模糊模式匹配的方法进行准确定位。实验结果表明该方法比在复杂背景中直接利用肤色信息检测人脸的方法速度快,准确率高,鲁棒性好。Based on visual perception,a face detection approach from coarse to fine matching in complex background is described in this paper.In the first stage,our approach approximately detects the image region where the probability of finding a face is high;during the second stage the fine matching is performed to verify the existence of the face in the candidate face region and accurate location of the verified face is performed.In the approximate detection,a stochastic model to characterize the skin color distribution of human skin is used to segment face and other skin areas from the background.In the fine matching,skin-color model,hair-color model and head-models of the different direction are used to perform accurate detection and marking.Experimental results show that this face detection method is faster,more accurate and robust than the face detection method by directly taking advantage of color information.

关 键 词：复杂背景 人脸检测 定位 肤色模型 色彩分割 模糊模式匹配 计算机视觉 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]