A New Face Recognition Algorithm Based on DCT and LBP

The paper proposes a new face recognition algorithm. Firstly, Discrete Cosine Transform (DCT) is conducted on an input face image. A few DCT coefficients on the left top corner are chosen as the global feature. At the same time, the image is divided into several parts. Local Binary Pattern (LBP)is conducted on each part and then LBP histogram sequences (Uniform LBP used) are accepted as the local feature. Secondly, the paper fusions the global and local feature using feature level fusion. Finally, Support Vector Machine (SVM) is adopted as the classifier and experiments are done on the ORL database. A result of 95.5% recognition rate and 5 ms time elapsed for each image is obtained, which shows the efficiency and practicability of the proposed algorithm and the correctness of feature level fusion. At last, the paper drops a conclusion that application of feature level fusion in face recognition will draw more and more attention in the future.