摘要: As the next generation of media, 3D video is attracting increased attention. a sequence three dimensional mesh models, captured and generated for real dynamic object. In this paper, we present simple framework motion editing in database to re-use data. Our system composed two modules. first module, automatically set up from original sequences off-line by analyzing feature vectors each frame. It observed that our have two-level temporal structure. A fine-to-coarse method proposed extract such segmented into fine-level structure reference frame strategy then clustered coarse-level second users can synthesize motions edit new online. cost function optimized transit between with users' requirements. All algorithms are based on analysis vector space edited played using OpenGL.