基于MPEG压缩域的快速、鲁棒的全局运动信息提取方法  

Efficient Global Motion Estimation Method for MPEG Video

作　　者：李宏亮[1] 刘贵忠[1] 李永利[1] 张忠伟[1] 

机构地区：[1]西安交通大学电子与信息工程学院,信息与通信工程系陕西西安710049

出　　处：《电子学报》2004年第1期121-124,共4页Acta Electronica Sinica

基　　金：国家教育部博士点基金 (No.2 0 0 0 0 6982 8);国家教育部骨干教师基金 (No .2 0 0 0GJ1 7)

摘　　要：全局运动信息在视频分析中起着重要的作用 .本文根据MPEG编码特点 ,提出了一种从MPEG压缩域中快速有效地进行全局运动参数估计的算法 .该算法充分利用了MPEG压缩码流中的信息 ,通过提取预测残差DC图像的运动背景区域 ,估计全局运动参数 ,从而保证了参数估计的准确性 ,有效地克服了已有文献中仅仅采用运动矢量进行全局运动估计的局限性 .根据不同的MPEG测试序列的对比分析 ,结果表明 ,本算法可快速准确地对MPEG视频序列进行全局运动信息估计 ,同时具有很高的鲁棒性 .Motion estimation plays an important part in the analysis of video sequences for the purposes of information extraction. In order to process video data effectively, camera motion extraction from compressed video is often required. An effective global motion estimation algorithm based on the moving background derived from MPEG compressed data is presented. The algorithm uses the predictive error DC-image to extract the moving background of the current frame, and then exploits a new global motion estimation method based on a 6-parameters motion model to calculate the camera parameters. In the evaluation experiments, five video test sequences with global motions and three video sequences without global motions are used to test the proposed algorithm, and compared with an existing method. The experimental results show our algorithm can estimate global motion parameters more effectively and exhibit more robustness than the existing algorithm.

关 键 词：镜头运动 背景估计 MPEG视频 

分 类 号：TN911.73[电子电信—通信与信息系统]