Creating 3D Personal Avatars with High Quality Facial Expressions for Telecommunication and Telepresence

This study aims at providing a low-cost solution for telepresence where people are reconstructed as 3D avatars using an ordinary webeam, while still exhibiting abundant facial information (such as micro-expressions) that are critical for face-to-face communication. We estimate the basic 3D shape and texture of the body from a set of video frames, and then subsequently update its body pose, facial expression, and facial texture in each frame. Our method is expected to reduce the entry barrier of VR systems and create an embodied telecommunication that conveys rich information and subtle emotional changes to deepen mutual understanding at a distance.