基于傅里叶的卡通动画形状上下文动作捕捉  

Cartoon Motion Capture with Shape Context Based on Fourier Descriptors

作　　者：李文博[1] 刘艳[1] 苏文瑛[1] 

机构地区：[1]天津大学计算机科学与技术学院,天津300072

出　　处：《计算机与现代化》2013年第3期168-171,共4页Computer and Modernization

摘　　要：对已有二维卡通角色进行动作捕捉,然后重新利用的传统方法效率很低。本文提出的方法可以在过滤掉源卡通角色中的冗余信息提高计算效率之后,捕捉更多角色信息以供重用。首先用基于YUV双阈值的方法得到角色的姿态序列(包含角色轮廓和内部线条信息),然后由傅里叶描绘子过滤冗余信息的同时保留所需要的角色动作信息(局部和全局形变),最后用形状上下文捕捉角色轮廓和关键点信息。实验结果表明,此方法可以高效实时地从源卡通角色中捕捉动作和形变信息以供重新利用。Motion capturing from existing cartoon character for cartoon reuse is time-consuming.This paper proposes a method to filter the redundant information in the source character and at the same time maximize both deformation and motion information obtained effectively.It consists of three steps： Firstly,the YUV double threshold based method is used to obtain a cartoon motion sequence including contour and drawing lines inside;Then,filter the redundant details and smoothen the contour by Fourier descriptors;Finally,capture the contour and key points from the sequence we have obtained by shape context.The experimental result shows that the motion and deformation information could be obtained more efficiently with our method.

关 键 词：卡通动画重用 背景减除 姿态序列提取 傅里叶描绘子 形状上下文 卡通运动捕捉 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]