三维人体动画的生成是计算机动画领域的热点问题。已有的运动捕获方法虽然可以获得逼真的运动数据,但使用不便、代价昂贵的缺点限制了其应用。利用已有的运动数据库,提出了一种视频驱动的三维人体动画合成方法。该方法通过建立运动库概率转移和统计模型,挖掘运动数据的时空结构,能够根据视频内容合成出新的动画数据,从而提高已有运动数据的可重用性,大大降低动画制作成本。采用广播体操运动视频验证了方法的有效性。
The generation of 3D human animation is a popular issue in the area of computer animation, Although the motion capture method can provide highly detailed motion data, it is too expensive and cumbersome for commercial use. A video-driven approach to 3D human animation with motion data was proposed, By building probability and statistical model of motion database, the spatio-temporal structure underlying the motion database is explored with the guidance of which 3D human animation that meets the content of video is synthesized. This method can make previous captured motion data reusable, thus reduce the cost of animation production greatly. The effectiveness of this method is verified by sport exercise video.