An Approach to Video Object Extraction in the Compressed Domain
基于压缩域的视频关注对象提取

In this paper, we propose a new method to segment video object. Firstly, we smooth each frame with the mean-shift and region-growing method. And then, the motion vectors which are released by video coding are used to extract a moving object. Finally, the object segmentation method proposed in this paper is used to extract the video attention object. Experimental results show that the proposed method can effectively separate the video object from the sequence and has the strong robustness in multi-targets tracking especially for switching targets and deformed objects.