一种基于内容表示的图像序列运动分割算法

A motion segmentation approach for content-based video representation in image sequence

提出了一种MPEG 4标准所规范的基于内容表示的图像序列运动分割算法。首先进行图像序列二维运动估计并计算位移帧差 ,然后 ,将位移帧差作为将运动物体从静止背景中分离出来的标准进行分割 ;最后对分割结果进行形态滤波。实验结果显示 ,在物体存在整体运动的情况下 ,分割结果表现出了很好的准确度 ,可以成功地提取出图像序列中的运动物体。

In this paper,a motion segmentation approach for content based video representation specified in the standard of MPEG 4 is proposed.By two dimension motion estimation and computing the displaced frame difference,the motion object can be separated form the still background,and then morphological filtering is used to gain the better segmentation result.Results obtained with MPEG 4 test sequences show that the accuracy of object segmentation is substantially improved in presence of global motion.