Study Of Face Recognition With Small Sample Size

With the development of science and technology, biometrics has became a key technology in the field of authentication.The biometrics technology includes:face recognition, fingerprint recognition, iris recognition, retinal recognition, speech recognition, shape recognition, signature recognition and so on. Among them, face recognition has obtained extensive attention, it has been widely used in many areas.Face recognition is a computer technology for authentication, it's an analysis and comparison of facial feature.In recently years, this technology have breakthrough development at home and abroad with the upsurge of research in face recognition.However, in some conditions there was small sample size(some others just one training image), so that the face recognition was decreased somewhat(some was unable to recognize). It is absolutely necessary to research this issue.In this paper, by systematically analyzing of relevent algorithms of face recognition, we present a novel method for face recognition with small sample size based on virtual information and wavelet-transform.In addition, we have research on Mirror-like Gabor features for face recognition under varying illumination conditions.The main research work of this article is as follows:(1)Analysis of the purpose and significance of the face recognition with small sample size, the research situation at home and abroad, theoritical and technical background for the presented method in this paper.(2)Discussed the viutual images, principal component analysis(PCA) and wavelet-transform.A method for face recognition with small sample size based on virtual information and wavelet-transform is proposed. Selected the optimum method to increase virtual samples.The send-low frenquency wavelet component was used to replace the original images.Experiment in ORL and AR person image database indicate the validity of the method.(3)Analysis of Gabor Filter and Mirror-like Gabor features for face recogition under varying illumination conditons.