Facial Parts Localization Based On Exemplars And Its Application In Face Recognition

Facial parts localization has become a hot technology in the field of face engineering, which is one of the most important branches of computer vision. As the growth of computer in the performance of hardware and the deeper research of relative algorithm, computer vision has gained lots of progress in recent years. Face recognition which is one of the most important branches of computer vision, can bring huge convenience into human being’s life. Facial parts localization is one of the key technologies in the field of face recognition, for its usability in both face alignment and face verification.This paper implements and improves the facial part localization system based on a library of exemplars. The facial part localization system based on a library of exemplars can overcome the shortcomings that both AAMs algorithm with strong constraint and DPM algorithm with weak constraint have. So our system can achieve a better accuracy. The fact that it doesn’t need a complex algorithm to apply the iterations or operations of searching makes our system more efficient either. The main efforts of our work are that:1. By surveying the main algorithm in the field of facial parts localization, we summarize their similarities and differences, and implement our algorithm by imitate their form;2. By applying some experiments on local features like HoG and SIFT, we select the feature which has better performance to implement our local detectors;3. We implement the global constraint for facial parts based on exemplars and improve it by adding a stronger global constraint which has the form of Markov Chain.4. By applying experiments between face verification systems with our framework and without, we prove the important role that our facial parts localization system plays.