一种支持视频检索的对象标志位编码算法

An Object Flags Encoding Algorithm Supporting Video Retrieval

针对传统标志位编码算法的编码代价大、海量监控视频难于浏览等问题，提出一种支持视频检索的对象标志位高效编码算法。基于对象区域信息、语义信息生成标志位来存储监控视频，根据视频编码和视频分析结果，生成对象区域信息与语义信息，利用基于区域生长的帧内编码算法消除空域冗余，基于运动估计的帧间编码算法从分像素精度上消除时域冗余，将对象区域信息、语义信息同时编码到原始视频码流中，实现监控视频的快速浏览与检索。实验结果表明，与基于欧式距离的帧内编码算法相比，该帧内编码算法的编码代价降低4%~14%，帧间编码算法降低28%~48%；基于对象标志位的视频解码能生成用户感兴趣的检索视频，提高用户浏览效率。

Aiming at the problem that traditional flags coding methods cost too much and the mass surveillance video is hard to browse,this paper proposes an efficient object flags coding algorithm supporting video retrieval. The algorithm generates the object flags to store the surveillance video based on object area information and semantic information. The object region information and semantic information can be generated according to the result of video coding and video analysis. In order to reduce the airspace redundant, a novel region information coding algorithm based on intra-frame region growing is proposed. And inter-frame object flags coding algorithm which is based on motion estimation eliminates the temporal redundancy further with sub-pixel precision. Then it writes the encoded object area information, semantic information into original video stream. Experimental results show that, compared with the coding algorithm based on Euclidean distance,the proposed intra-frame object flags coding algorithm makes the coding cost reduced by 4% ~14% , and inter-frame algorithm is reduced by 28% ~48% . Video decoding based on object flags can improve the efficiency of the user to browse by generating user interested video.