【作者】 张小红； 唐敏； 蒋志成； 童若锋；

【Author】 Zhang xionghong,Tang min,Jiang zhicheng, Tong ruofeng (College of computer science and Technology,Zhejiang University,Hangzhou 310027,China)

【机构】 浙江大学计算机科学与技术学院；

【摘要】 针对前景视频对象的运动信息对视频融合结果真实感的影响,提出了一种基于视频对象轮廓曲线的曲率特征匹配的视频对象运动信息消除技术。通过提取对象视频的轮廓曲线,在尺度空间计算轮廓曲线的曲率并进行曲率特征匹配,从而得稳定得到块运动估计矢量;我们在计算全局运动矢量之前对块运动估计矢量进行过滤,提高算法的可靠性。实验表明,该算法能有效消除视频对象的运动信息。更多还原

【Abstract】 The movement of foreground video object will affect the sense of reality of video composition result. This paper proposes a method of removing video object motion information based on curvature feature matching of the video object’s outline curve.By extracting the outline curve of the video object and computing the curvature of the outline curve and then matching the curvature features,we can get the block motion vector; Before computing the global motion vector,we improve the accuracy rating of the block motion vector by a filtering process.The experiment results show this method can efficiently remove the movement of video object.personalized human models.First a series of human更多还原