Probabilistic Graphical Model for Multiple Facial Feature Tracking

Tracking multiple facial features simultaneously is a challenge when rich expressions are presented on a face.Several independent condensation-style particle filters are utilized to track each facial feature in temporal domain.Particle filters are very effective for visual tracking problems,however multiple independent trackers ignore the natural relationships among facial features.We use Bayesian inference-belief propagation to infer each facial feature's contour in spatial domain,taking into consideration the previously extracted relationships among contours of facial features which are organized as a large facial expression database.Experimental results show that our algorithm is robust.