Animation of 3D characters from single depth camera
Mian Ma, Feng Xu, Yebin Liu · 2011
As the development of depth acquisition techniques, in recent years, depth cameras achieve to obtain depth maps in real time. In this work, by using a single depth camera, we propose a method to transfer 3D motion of a performer to another character (avatars or other human characters). We capture the motion of a performer by a Kinect camera, which outputs the 3D positions of 15 joints of the performer for each frame. Even though the joints positions are quite noisy and with errors in some frames, our method achieves to calculate joints' rotations with spatial and temporal consistency and transfer the motion to a target skeleton. By using a skinning technique, the captured skeleton motion drives the target 3D model which performs the same motion with the captured performer.