Face recognition using processed histogram and Phase-Only Correlation (POC)

This paper proposes an automatic face recognition technique using mutual combination of bin-based histogram processing and Phase-Only Correlation (POC) - based techniques to obtain effective recognition accuracy. The first phase preprocesses the faces with bin-based histogram approach and extracts the preliminary results which are rotation invariant in nature. The second phase verifies the face patterns with POC-based matching technique in order to obtain the exact results which are the same not only distribution-wise but also content-wise. The effective preprocessing by bin-based histogram approach filters out the mismatches and helps in reduction of the overall complexity of POC. The empirical results obtained on ORL database shows the recognition accuracy of 99.5%, which is very promising and is comparable with any other face recognition scheme.