Face recognition using extended vector quantization histogram features

Face recognition is a typical application of biometrics identification technologies, which requires specific methods to obtain face representation as its features. In this paper, we apply a simple yet highly reliable method called Vector Quantization (VQ) to extract the features. Although VQ algorithm has been proven effective, the inability of VQ histogram features to convey spatial structure information, however, greatly limits its discriminating capability. So in this paper, we propose a novel framework called Markov Stationary Features (MSF) based on selected direction which can not only encode the spatial structure information into VQ histogram but can also eliminate the inherent ambiguity of the features extracted from the facial image so as to realize the goal of improving the face recognition performance. Experiments are conducted by using ORL face database and the maximum average recognition rate of 96.28% can be obtained. By combining multiple MSF-VQ features based on different directions, the recognition rate can increases up to 96.45%.