The recent success of the Kinect sensor has a significant
impact on 3D data based computer applications. This study
aims to obtain MPEG-4 compliant realistic and animatable
face models from Kinect video. The complete framework for
this process includes initially the computation of high quality
3D scans from RGB-D Kinect video, and then the
computation of animatable MPEG-4 face models using these
high quality scans. This study shows that it is possible to
obtain high quality 3D scans and realistic and animatable
face models of subjects using lower quality Kinect data.
Index Terms- Kinect, MPEG-4, animatable face model