基于图论的人脸图像数据降维方法综述

摘要

With the method based on graph theory has gotten more and more attention in recent years,to reduce the high dimensionality of data is the core problem in face recognition.The paper introduces the basic concept of graph theory,summarizes different kinds of methods about dimensionality reduction for face image,which can be unified to graph embedding framework.Then it analyses the advantages and disadvantages of all kinds of algorithm in the linear and nonlinear respect,and concludes that the mathod of nonlinear graph embedding has better function in both nonlinear respect and data dimensionality reduction than traditional ways.Finally,the paper discusses the development in the future according to the existing ways of graph construction.%近几年基于图论的降维方法越来越得到人们的关注,本文针对人脸识别中的核心问题即对高维数据进行降维的目的,首先介绍了有关图论的基本概念,通过总结各种人脸图像降维的方法,将这些方法统一到图嵌入框架中.然后结合线性与非线性的角度分析了各种算法的优缺点,得出了非线性图嵌入算法在挖掘人脸图像中的非线性特征以及在数据降维方面均优于传统的方法.最后针对现有的构图方式所存在的问题对今后的研究与发展方向进行了讨论.