Face Recognition Methods Study Based On Subspace Analysis

Face recognition has become a hot research in the biometric recognition,because it is natural and friendly.People's identifying information can be recognized by face recognition, then guarantee information security.Due to these features,face recognition has been applied on many fields,such as national security,information security,and it has high scientific value and commercial value.Subspace analysis is the focus of research in face recognition field.Comparing with other technologies, subspace analysis technology has several advantages such as high recognition rate, strong robustness,less calculative quantity.This thesis studies face recognition study base on subspace analysis technology.This thesis studied the face recognition technologies,contrasted and summarized advanages and disadvanages of subspace analysis technology.Based on of present literatures,this thesis improves two algorithms,including combining with image fusing base on subpace analysis algorithm and combining with image pyramid transform base on subpace analysis algorithm.Firstly,this thesis studies application of wavelet transform in the field of face recognition.The image fusion technique is adopted in face recognition field.It could synthetic a new sub-image from same targets in three high frequency sub-images. Applying dual-subspace feature extraction algorithm can overcome the limits of single subspace.So this thesis combines Principle Component Analysis(PCA) with Non-negative Matrix Factorization (NMF)to improve the rate of face recognition. And the algorithm computes the contributions made by image entopy.The algorithm can improve the rate of face recognition.Secondly,this thesis applies Laplacian pyramid transform into face recognition field.The image was decomposed into three-level sub-image by using Laplacian pyramid transform.Then the method computes the contributions made by image entopy.The amount of principle component does not decide.So this thesis studies the problem by using PCA and Two Dimension Principle Component Analysis (2DPCA) methods.The experimental result shows that the algorithm can improve the rate.And the rate base on 2DPCA combine with Image pyramid transform is higher than base on PCA.The innovation is as follows:This thesis present an algorithm that combine with image fusing base on PCA and NMF methods.The algorithm not only makes the best of features that extract from high frequency sub-images,but also complements PCA and NMF methods.So the rate is improved.This thesis present an algorithm that combine with image pyramid transform base on subpace analysis.The algorithm use PCA and 2DPCA to prove its efficiency.The experimental result shows the amount of principle component.