Study On Image Representation And Classification For Face Recognition

Face recognition is one of the most important research topics in pattern recognition,machine learning,computer vision,image analysis and understanding.In the past few decades,face recognition has made remarkable progress,but still suffers from some bottlenecks of the robust image representation,the efficient similarity metric and so on.In this dissertation,we focuse our attention on the creative issue of theory,model and algorithm of robust face representation and classification.The primary works and remarks are as follows:Inspired by the subspace properity of singular value decomposition for face image,this paper presents a simple but effective subspace based method for face recognition,named nearest orthogonal matrix representation(NOMR).Specifically,the specific individual subspace of each image is estimated and represented uniquely by the sum of a set of basis matrices generated via singular value decomposition,i.e.the nearest orthogonal matrix(NOM)of original image.Then,the nearest neighbor criterion is introduced for recognition.Compared with the current specific individual subspace based methods(e.g.the sparse representation based classifier,the linear regression based classifier and so on),the proposed NOMR is more robust for alleviating the effect of illumination and heterogeneous(e.g.sketch face recognition),and more intuitive and powerful for handling the small sample size problem.Generally,the conventional sparse representation based methods treat the face image as the linear superposition of identity imformation and noise.However,Lambertian Reflectance Model and amount practical examples have shown that it is not always the case,especially in the case of illumination changes comprised.Rather,in this paper,a new sparse representation based scheme,illumination sparse representation based classifier(ISRC)is developed to overcome the weakness of the conventional sparse representation based methods as applied in illumination variation face recognition.In the proposed method,the identify information and the illumination will be separated and obtained by an optimizing model simultaneously.Experimental results on several face database show the proposed ISRC is more effective than conventional sparse representation based methods such as SRC,CESR and RSC.A novel image similarity measurement method named transformation similarity measurement(TSM)is proposed for face recognition.The TSM calculates the degree of similarity by the liner transformations between two images.This paper provides a detailed analysis about the reasonable of the TSM and demonstrates its advantage features in face recognition.Based on the TSM and nearest neighbor rule,a specific algorithm,named TSM-NN,is proposed for face recognition.The proposed method was tested and evaluated using several face databases:Extended Yale B,AR,CMU-PIE,FRGCv2 and CUHK Face Sketch database(CUFS).Experimental results demonstrate that the proposed method achieves encouraging performance compared with the state-of-the-art methods.The linear reconstruction measure(LRM),which determines the nearest neighbors of the query sample in all known training samples by sorting the minimum l2-norm error linear reconstruction coefficients,is introduced in this paper.The intuitive interpretation and mathematical proofs are presented to reveal the efficient working mechanism of LRM.Through analyzing the physical meaning of coefficients and regularization items,we find that LRM embodies collaborative measure mechanism and provides more useful information and advantages than the conventional similarity measure model which calculates the distance between two entities(i.e.conventional point-to-point,C-PtP).Inspired by the advantages of LRM,the linear reconstruction measure steered nearest neighbor classification framework(LRM-NNCF)is designed with eight classifiers according to different decision rules and models of LRM.Evaluation on several face databases and the experimental results demonstrate that these proposed classifiers can achieve greater performance than the C-PtP based 1-NNs and competitive recognition accuracy and robustness compared with the state-of-the-art classifiers.