Face Recognition Using Principal Component Analysis of the Wavelet Packet Decomposition
Vytautas Perlibakas · SSRN Electronic Journal · 2004
Principal Component Analysis (PCA) or Karhunen-Loeve transform (KLT) - based face recognition method was investigated by many researchers and usually it is noted that usage of the PCA-based face recognition method with large face databases is not practical, because using this method we need to calculate eigenvectors and eigenvalues of the large covariance matrix. In order to solve this problem we can use smaller number of training images and covariance matrix decomposition, incremental eigenspace learning, hardware implementation of the KLT. In this article we propose a novel face recognition method based on the Wavelet Packet Decomposition (WPD) and Principal Component Analysis (PCA) and show that using the proposed method we can reduce the time-complexity of the classical PCA-based face recognition method and achieve similar face recognition accuracy. The proposed method could make the PCA-based method more suitable for face recognition with large databases, because its training time is independent from the number of training images. Using the proposed face recognition method we achieved 82-89% first one recognition with the database containing photographies of 423 persons.