Output details
34 - Art and Design: History, Practice and Theory
Bournemouth University
Mean value coordinates–based caricature and expression synthesis
Originality: Creating plausible animated faces is still a challenging task for visual effects professionals. One approach to this problem is to extract the facial expression from a video clip or from a 3D scan, which is then transferred to a target character. Our work presents a novel method for the synthesis of facial expressions and caricatures based on mean value coordinates (MVC). Facial expression transfer can be applied to both 2D and 3D face models. The resultant editing algorithms allows the manipulation of the facial expressions, including exaggeration.
Rigour: To extract the style of a specified caricature face or facial expression, our algorithm employs machine learning techniques that can effectively extract expression features based on a training dataset. Experiments have demonstrated that our algorithm can be applied to any single frontal face images for caricature or expression synthesis.
Significance: This work is a continuation of our on-going research on multimodal face recognition and animation. It provides modellers, animators and caricature artists a versatile tool for creating different visual artefacts of the human face, allowing a traditionally tedious work to be done both intuitively and efficiently.