提出了一种基于递归最短生成树算法的H.264压缩域实时分割运动对象的算法。首先将从H.264编码端提取的运动矢量进行归一化、空间内插，得到稠密运动矢量场，再采用全局运动补偿技术抵消全局运动的影响，最后采用改进的“递归最短生成树”（RSST）算法对稠密运动矢量进行聚类，实现对运动对象的分割。实验结果表明，该算法对视频序列能实现较准确的分割。

In this paper, we present a novel approach to segmenting moving objects from the H.264 compressed based on recursive shortest spanning tree(RSST). Firstly, the motion vector is obtained from the H.264 encoder and then a dense vector field is formed after the vector is unified and accumulated with a backward iterative projection process. Secondly, global motion compensation is used to eliminate the global motion. Thirdly, a modified RSST arithmetic is used to make the blocks clustered.Thus,with the proposed method we achieved moving object segmentation. Experimental results for several video sequences demonstrate the effectiveness of the proposed approach.