Interactive facial sketch expression generation using local constraints

In this paper, an interactive sketch expression generation method is presented. By dragging one or a few control points on a new input sketch face, the system can draw corresponding natural expressions automatically. However, for an unseen target, with just a few knowledge of its neutral face, users' arbitrary input will often lead to unnatural even unreasonable results. In the proposed method, two kinds of priors are learnt from a set of different facial examples first-facial shape prior and correlated motion prior, which guarantee the global facial structures and the specific correlated motion regulations of the face respectively. Then, we present an integrated statistic face model to combine these two priors with user-input local constraints. The experimental results showed that our method could generate natural expression results to satisfy the user's requirements. Our interactive system is also easy and simple to operate even for the first-time user.