18-3 Motion Structure Extraction for 3D Video Editing

抄録

3D video is captured by multiple cameras and expressed by mesh models. 3D video generation is very time-consuming and expensive. Therefore, it is necessary to edit 3D video to re-use it. In this paper, we propose a motion structure extraction method, which is the prerequisite of 3D video editing. It is observed that our 3D video sequences have a multi-level motion structure. By analyzing the feature vectors in each frame, the multi-level structure is extracted with a Harming window. The experimental results show our method is effective.