Face verification using multiple localized face features

Abstract

In this paper, we described face verification execution using different feature extraction methods on different regions of the face. We chose two feature extraction methods which are the Discrete Cosine Transform (DCT) and the Local Binary Pattern (LBP). Classification is done separately on for each region using Support Vector Machine. The final verification decision is calculated by combining the classification scores of the face region. The results show that generally, LBP gives better results than DCT on our dataset and parameter settings but both methods did not extract good discriminating features on the nose region. © 2013 Springer International Publishing.