Efficient Reconfigurable Architecture to Extract Image Features using Local Binary Pattern for Face recognition

Abstract

Abstract Recognition of the face is a widely held method to detect human features. In various claims, like detecting criminals over video surveillance, face recognition is the most practical and computable recognition technique for humans. In this paper, we propose Efficient Reconfigurable architecture to extract image features using Local Binary pattern (LBP) for Face Recognition. The face image is preprocessed using Gaussian filter to remove high frequency components. The preprocessed image is then applied to optimized LBP block to obtain the LBP features followed by Histogram. The counter and comparators are used along with moving window architecture which leads to less complex LBP architecture. The Histogram LBP features obtained for both database sample and test sample are further compared to make the decision for recognition. The simulation is performed for Olivetti Research Laboratory (ORL) dataset using MATLAB by showing FAR, FRR and TSR values. Further, the proposed architecture is synthesized on Spartan 6-xc651x4c-3csg432 Digilent FPGA board. It is observed that the recognition time of our architecture is 1.05 µS which is better compared to existing methods.