Abstract
In this thesis, we propose a new system for capturing animation by detecting character motions in a single video stream We track the motion by using the skeleton model that animators provide and retarget the motion onto 2D drawings. By using video sequences as the source, a new animation is created with a different look but will imitate the movement of the original character in the video. The animation imitate process is done through first track the motion by calculating the optical flow of succeeding frames, for this step we will ask animators to specify all the feature points and join the feature points to be a 2D skeleton model with hierarchical structure. We then apply a shape deformation method that will keep the shape rigidity of the original 2D retargeting character. By combining these existing methods, our system can produce rough animations that an animator can then further modify the animations and do some additional refinements.