视频运动特征提取模块的实现

Realization of visual motion feature extraction module

基于内容的视频检索为人们检索具有相似内容的视频数据提供了新的手段,而运动信息作为视频内容中的一种特有信息,是视频检索领域研究关键问题之一。通过对运动特征提取算法进行研究,设计并实现了一个实用的全局运动特征和局部运动特征提取模块。实验表明:该模块能够有效地分割全局运动与局部运动,提取的运动特征信息可作为基于内容的视频相似检索系统的重要索引。

Content based video retrieval （CBVR） provides new ways to retrieve similar video. As a unique type of information, motion feature has its special importance to the performance of a video retrieval system. By study of the existing extraction and representation algorithm of motion information, a practicality module is designed and implemented mainly on global and local motion feature extraction. The expermental results show that the module can segment global and local motion feature efficiently and extract index feature for the whole video similarity retrieval system.