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

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.