Discriminative and Generative Model Learning for Video Object Tracking