基于三维多分辨率模型与Fisher线性判别的人脸识别方法  

Face Recognition Based on a 3D Multi-Resolution Model and FLDA

作　　者：谷春亮[1] 尹宝才[1] 孔德慧[1] 胡永利[1] 

机构地区：[1]北京工业大学多媒体与智能软件技术实验室,北京100022

出　　处：《计算机学报》2005年第1期97-104,共8页Chinese Journal of Computers

基　　金：国家自然科学基金 (60 3 75 0 0 7);北京市自然基金重点项目基金 [4 0 110 0 1(KD0 70 62 0 0 10 1) ];北京市基金委重点项目基金[KZ2 0 0 3 10 0 0 5 0 0 2 (KP0 70 62 0 0 3 74) ];重点实验室开放基金 (KP0 70 62 0 0 3 71)资助 .

摘　　要：提出了一种不同姿态和光照条件下的人脸识别方法 ,将三维多分辨率模型与Fisher线性判别结合起来 .为了排除光照、姿态对人脸识别的影响 ,利用重采样技术构造了三维多分辨率模型 ,更快、更精确地提取人脸特征 ;同时结合Fisher线性判别 ,充分利用不同条件下的二维人脸图像信息 ,更有效地排除光照、姿态的影响 .实验表明 ,三维多分辨率模型与Fisher线性判别相结合能够很好地适应外部条件的变化 ,提高了人脸识别的速度和效率 .A new method of face recognition across different poses and illuminations is proposed in this paper, which combines a 3D multi-resolution model and Fisher's Linear Discriminant Analysis (FLDA) together. Analysis of poses and illuminations is a very difficult work, in order to get rid of extrinsic effects, the method is based on a 3D multi-resolution face model that is formed on the basis of gridded-resampling and can encode shape and texture in terms of model parameters fleetly, and an algorithm that recovers these parameters from a single image of a face is proposed. Fisher's Linear Discriminant Analysis is applied to make full use of images acquired from various conditions. It can make recognition independent of imaging conditions more efficiently. The results show that combination of the multi-resolution model and FLDA can adapt to variety of poses and illuminations, and get a considerable effect.

关 键 词：人脸识别 三维多分辨率模型 FISHER线性判别 重采样 匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]