Real-time discrimination of frontal face using integral channel features and Adaboost

In this paper we present a novel approach for discrimination of frontal face in video, using integral channel features(ICF) and Adaboost. We have two stages for this approach based on classification, the first stage is training process, we utilize ICF exacted from training database to train strong classifier, which is implemented by Adaboost. The second stage is discriminating process by scoring, we compute ICF of the face detection window, and then scoring the window using the trained classifier, and at last the most frontal face will be chosen by the highest scores. Furthermore, we then apply the approach to the ChokePoint database and compare with different approaches, showing a good performance.