Compact binary patterns (CBP) with multiple patch classifiers for fast and accurate face recognition

Face recognition is one of the most active research areas in pattern recognition for the last decades because of its potential applications as well as scientific challenges. Although numerous methods for face recognition have been developed, recognition accuracy and speed still remain a problem. In this paper, we propose a novel method for fast and accurate face recognition. The contribution of the paper is three folds: 1) we propose a new method for facial feature extraction named the Compact Binary Patterns (CBP), which is a more compact and efficient generalization of Local Binary Patterns. 2) We show that Whitened Principal Component Analysis (WPCA) is a simple but very efficient way to enhance CBP features. 3) To further improve the recognition rate, we divide a face into patches and perform recognition using multiple classifiers, whose weights are estimated by a Memetic Algorithm. Our method is tested thoroughly on the FERET dataset and achieves promising results.