基于动态规划的关键帧提取算法

在传统的关键帧动画中,动画师对关键帧进行编辑并通过插值生成一段连续的动画.运动捕捉数据的每一帧均可看作关键帧,但是对其直接进行编辑非常繁琐,需要从运动中提取具有代表性的关键帧,简化运动的编辑操作.已有的关键帧抽取方法主要基于一些局部误差度量策略,算法的全局误差难以得到保证.提出了一种基于动态规划的运动捕捉数据的关键帧提取算法,该算法通过曲线拟合技术对密集采样的运动捕捉数据进行减帧,在生成指定数目关键帧的同时保证误差达到全局最小.用户可以利用传统的关键帧技术对减帧后的运动进行编辑.

Dynamic Programming Based Key-frame Extraction Algorithm

In traditional key-frame animation,animators edit the key frames to generate a continuous animation sequence by interpolation.However,each frame of motion capture data can be considered as a key frame,and it is too tedious to perform editing task to such data.We need to extract the representative key frames from motion to simplify the motion editing operation.Existing key frame extraction algorithms mainly consider local error metrics,thus do not guarantee a global error tolerance.In this paper,we propose a dynamic programming approach to key reduction for motion capture data,which can reduce key frames of intensive sampled motion capture data through a curve fitting algorithm,and this method promises a minimized global error with a given number of key frames.In addition,users can edit key-reduced motion capture data using traditional technologies of key-frame.