Motion Structure Parsing and Motion Editing in 3D Video

3D video, which is captured by multiple synchronized cameras and stored in mesh models, is emerging in recent years. However, the generation of 3D video is time-consuming and expensive. In this paper, we present an editing system to re-use 3D video efficiently. The hierarchical motion structure in 3D video is observed and parsed. Then, the representative motions are selected into a motion database, where the user can choose the desired motions. When synthesizing those chosen motions, the motion transition is optimized by a cost function. Some other information is also displayed in the interface to ease the editing. It should be mentioned that all the analysis and processing in our system are done in feature vector space.