Abstract: To solve the problem that the segmentation accuracy of moving object is not high enough under complex background in compressed domain video, this paper proposes a moving object segmentation method based on the up-to-date video coding standard HEVC. Firstly, block partitions and corresponding motion vectors extracted from the HEVC bitstream are classified based on spatial and temporal consistency in intra frame and inter frames. Then the MRF model is selectively exploited to refine the classification result for more accurate video object. Experimental results show that the proposed method can achieve an effective and reliable segmentation result, and the method is better than the other comparison methods especially for multiple target motion video segmentation.

Key words: High Efficiency Video Coding（HEVC）, moving object segmentation, block partition, Markov Random Field（MRF）

摘要： 针对压缩域视频的运动对象分割在复杂背景下分割精度不高的问题，提出一种基于最新压缩编码HEVC的运动分割方法。首先从HEVC压缩码流中提取块划分和相对应的运动矢量信息，并分别在帧内和帧间对运动矢量进行空域和时域的标签分类，然后利用MRF模型对标签场进行运动一致性估计，得到更精确的运动目标，最后输出MRF分割后形成的掩模信息。通过实验证明，该运动分割方法能够达到有效并可靠的分割效果，尤其对于多目标运动的视频分割效果优于其他比较的方法。

关键词: 高效率视频编码（HEVC）, 运动目标分割, 块划分, 马尔可夫随机场（MRF）