计算机人脸识别技术综述

概述了计算机人脸识别技术的历史及发展现状,讨论了在计算机人脸识别领域占有主流地位的Eigen脸方法(主元素分析方法)、最佳鉴别矢量集法(基于Fisher线性判别准则方法和基于Foley-Sammon变换方法)、Bayesian脸方法、基于傅里叶不变特征法和弹性图匹配法。指出了各个研究方向人脸识别方法, 给出了计算机人脸识别性能评价指标,包括识别率、计算时间、数据存储量和可扩展性等。根据这些性能评价指标,对当前的各种计算机人脸识别技术进行分析评价。讨论结果表明,基于Fisher线性判别准则的最佳鉴别矢量集法,Bayesian脸方法和基于傅里叶不变特征法都有较好的性能,具有一定的应用前景.

Survey in computer recognition technologies of faces

The history and the current status of computer recognition technologies of faces are introduced briefly. Then, several dominate methods using Eigen face (PCA), optimal discriminant vectors (based on Fisher criterion and based on Foley-Sammon transformation), Bayesian face, invariant Fourier features and elastic graph matching are discussed. Meanwhile, the principles and the newest states of these methods are given. Follow which are some computer recognition technologies of faces in other direction. At last, we give some performance evaluation criterion such as ratio of correct recognition, time of computing, quantities of data storage and performance of expanding, and according to those performance evaluation criterion analyses and evaluate a variety of computer recognition technologies of faces. Discussion show that based on Fisher criterion method, Bayesian face method and based on invariant Fourier features method give better performance, and can be practicable.