INTEGRATION OF LOCAL AND GLOBAL FEATURES FOR FACE RECOGNITION

This paper proposes a face recognition method that fuses information which is obtained from global and local approaches for improving performance under the pose and expression changes. The magnitude or phase information of Log Gabor transformed image is extracted by local projection entropy method. At the same time, the PCA performed on origin image is exploited to compensate for the loss of the global information in the local approach. Recognition is accomplished by fusing scores from both global and local approaches using weighted sum rules. Performance of the proposed algorithm is validated on public ORL and Yale Face database.