Caricature Drawing System as Understanding User Facial Attributes

Abstract

We have developed a system which generates a 2D caricature face of a user by understanding his facial attribute. In recent, there have been many approaches to make a user caricature face. They, however, analyzed a face with a 2D sparse facial feature alignment method and it caused lack of his facial attribute information. In this research, we tried to automatically analyze the facial attribute of a user using a bilinear face model, enabling to make the caricature reflecting the user facial attribute. It also provides to change the facial expression of the caricature intuitively since it is parameterized by identity and expression parameters. When user gives his face image to the system, it reconstructs his 3D face model automatically without any user input. The result 3D face model is expressed as identity parameter. We mapped the identity parameter to facial attribute parameter space, which makes the system notice what kind of notable facial attributes the user face has. As emphasizing or reducing the expression power of remark facial attributes, the system makes the 3D face model reflect the facial attribute of the user dramatically. It also enables the user to change the facial expression of his caricature as changing the expression parameter values. As the 3D model of the user caricature is projected to the 2D space with a cartoon-like texture effect, user can get his own traditional caricature drawing.