Face recognition algorithm based on the wavelet approximation coefficients and non-parametric discriminant analysis
Dengyi Chen · Journal of Beijing Information Science & Technology University · 2011
According to the wavelet transform theory,a face image is decomposed into the three-level wavelet coefficients by using the db1 wavelet function.Subsequently,a new sample vector is obtained by reshaping the three-level wavelet approximation coefficients.Because the non-parametric discriminant analysis has a good adaptability for the non-Gaussian distributed samples set,a face recognition algorithm based on the wavelet approximation coefficients and non-parametric discriminant analysis(WANDA)is presented in the paper.The non-parametric discriminant analysis is applied to the above-mentioned samples to form the between-class scatter and the within-class scatter matrices,and the Fisher's linear discriminant is used for face recognition.The experiment results show that the face recognition rate is respectively 95% and 97.5% under the ORL and the CAS-PEAL-R1 face database by the above algorithm.