Face recognition based on ordinal correlation

In this paper, we propose a new face recognition system based on the ordinal correlation principle. First, we explain the ordinal similarity measure for any two images and then propose a systematic approach for face recognition based on this ordinal measure. In addition, we design an algorithm for selecting a suitable classification threshold via using the information obtained from the training database. Experimentation is conducted on the Yale datasets and the results show that the proposed face recognition approach outperforms the Eigenface and 2DPCA approaches significantly and also the threshold selection algorithm works effectively. Further, we carry out experimentation with various noise algorithms and the results show that the ordinal approach outperforms the Eigenface and 2DPCA approaches under different noise algorithms. © 2007 Inderscience Enterprises Ltd.