姿态变化估计策略下的人脸真实性判别研究  

Face liveness verification using estimation of face pose variation

作　　者：任勇[1] 张楠[2] 牟轩沁[2] 

机构地区：[1]重庆大学通信工程学院,重庆400044 [2]西安交通大学电信学院图像处理与识别研究所,西安710049

出　　处：《计算机工程与应用》2009年第16期193-196,共4页Computer Engineering and Applications

摘　　要：针对人脸识别系统中的欺骗手段,提出了一种基于姿态变化的脸部真实性判别算法。在构建的真实性判别系统中,引入了新的脸部特征点匹配标准函数及变形块匹配方法,对脸部序列图像特征点对位置进行精确提取和进行运动估计,最终通过对脸部姿态变化的策略分析完成真实性判别。算法分析及实验结果表明,该算法特别对照片欺骗手段具有良好的防范作用,可以较好地加强人脸识别系统的安全性。Aiming at photo deception in face recognition system,a new face liveness verification algorithm using estimation of face pose variation is presented.A block match algorithm based on variable-block and the new matching criterion function were used to get accurate positions of facial feature points and motion estimation of facial feature points.The verification was achieved through analysis of pose variation and head movement.Algorithm analysis and experimental results show that the algorithm can distinguish live faces from attacks of still photoes ,and enhance the safety of the whole face recognition system .

关 键 词：人脸识别 真实性判别 人脸姿态 块匹配 运动估计 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]