Real-Time Face Recognition with SIFT-Based Local Feature Points for Mobile Devices

We present a fast and light-weight face recognition algorithm using local feature points for mobile device. To recognize face accurately, we adopt Gabor-LBP histogram and SIFT-based local feature point. Gabor-LBP histogram is used to represent the local texture and shape of face images. SIFT-based local feature point is used to select some regions which have high probability to contain more important information of face components (eye, nose, mouth, etc.). The training stage of the proposed method is similar to other face recognition algorithms based on LBP histogram. The proposed algorithm has the advantage in test stage. Only selected blocks are used in the test stage. The selected blocks contain one more local feature points extracted by SIFT detector. Comparison between gallery image (train image) and probe image (test image) performs Gabor-LBP histogram sequences of selected blocks. Therefore the proposed algorithm has merits in the aspect of processing time and memory. Experimental results show that the proposed method can be achieved a similar recognition performance with general face recognition algorithm using all blocks of face image. The proposed method has an outstanding performance in processing time and memory. It is suitable for real-time face recognition in mobile device.