基于姿态和光照不变的三维人脸识别研究  

3D Face Recognition Based on the Changeless Pose and Illumination

作　　者：张怡[1] 周渊[2] 

机构地区：[1]兰州工业高等专科学校软件工程系,甘肃兰州730050 [2]兰州交通大学电子与信息工程学院,甘肃兰州730070

出　　处：《河西学院学报》2011年第5期55-60,共6页Journal of Hexi University

摘　　要：为了解决由姿态和光照变化引起的误识率,提出了一种利用稀疏统计形变模型,进行三维人脸形状重建,通过调整角度,解决人脸图像姿态发生改变的方法.运用球面谐波商图像的方法,估计原始输入图像的光照属性,对生成的模型进行标准化的姿态和光照重构,解决了原始输入图像中由于姿态和光照不同所引起的识别误差.实验表明,该方法能够有效地提高对三维人脸的识别率.In order to solve the change caused deterrent rate by posture and illumination ,using sparse statistics to reconstruct the deformation model of 3D face shape. Through adjust the angle, solve the pose of changed face image method. Using spherical harmonic business image. The method of original input image of illumination estimates for the created model attribute, standardized posture and illumination reconstruction, solve the raw input image due to different attitude and light caused by the recognition error. The experiment results show that the method can effectively improve to the 3D face recognition rate.

关 键 词：三维人脸识别 姿态 光照 球面谐波 商图像 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]