Extended facial expression synthesis using statistical appearance model

Statistical model based facial expression synthesis methods are robust and easier to be used in real environment. But facial expressions of human are very various. How to represent and synthesize expressions which is not included in training set is an unresolved problem in statistical model based researches. In this paper, we propose a two step method. At first, we propose a statistical appearance model, the facial component model, to represent faces. The model divides the face into 7 components, and constructs one global shape model and 7 local texture models separately. The motivation to use global shape + local texture strategy is the combination of different components can generate much more kinds of expression than training set have and global shape guarantees to generate 'legal' result. Then a neighbor reconstruction framework was proposed to synthesize expressions. The framework estimates the target expression vector by linear combine of neighbor subject's expression vectors. This paper primarily contributes three things: First, the proposed method can synthesize a wider range of expressions than the training set have. Second, experimental demonstrate that FCM is better than standard AAM in face representation. Third, neighbor reconstruction framework is very flexible. It can be used in multi-samples with multi-targets and single-sample with single-target applications.