No abstract

This paper describes a facial animation playback scheme based on simultaneous encoding of captured face textures and geometric positions. The high quality facial data is acquired using a more robust variant of the Universal Capture (UCap) technique that incorporates state-of-the-art marker-based motion capture. A novel encoding, based on Principal Component Analysis, is used to encode the data and uses more components in areas of the face that require them yielding finer control over the visual quality of the reconstruction. Furthermore, a frame interpolation technique is demonstrated that operates within this component sub-space and provides smooth transitions between facial expression clips allowing us to effectively apply a motion graph approach to facial animation. The method runs in real-time for tens of high-fidelity characters on current hardware.

No abstract