Face recognition using vector quantization codebook space information processing

We propose a novel information-processing algorithm called Vector Quantization (VQ) codebook space information processing. Based on this algorithm, we have developed a very simple yet highly reliable face recognition method called VQ histogram method. Codevector referred (or matched) count histogram, which is obtained by VQ processing of facial image, is utilized as a very effective personal feature. Experimental results show recognition rate of 95.6% for 40 persons' 400 images of publicly available AT&T database containing variations in lighting, posing, and expressions. By combining multiple low pass filtering procedures, recognition rate increases up to 97% or higher. For utilizing the geometric information of the face, furthermore, a region-division VQ histogram method is proposed in this paper. We divide the facial area into 5 regions of facial parts (forehead, eye, nose, mouth and jaw), recognition results with different parts are first obtained separately and then combined by weighted averaging. Topi recognition rate of 100% is obtained by using the private database, which was taken in practical environment. Based on the VQ histogram method, moreover, we have also developed simple face recognition method called Adjacent Pixel Intensity Difference Quantization (APIDQ) Histogram Method, which is much simpler and has equivalent performance in feature extraction as compared with the VQ histogram method. By utilizing the table look-up (TLU) method in the quantization step, the total recognition processing time is reduced to only 31 msec.