Abstract:

A video retrieval scheme based on global motion is proposed. A detailed global motion analysis for Moving Picture Export Group（MPEG）videos has been made and a camera motion model has been extracted. The model parameters are used as a motion features sequence, and the query by example of retrieval scheme is based on sequential match. Furthermore, according to global motion parameters motion analysis functions are constructed to segment exactly visual global motion and annotate semantic information, and then, the eXtensible Markup Language(XML) is used to describe the video global motion to realize the semantic query by Xquery engine. The experiments show that in the scenery videos, this scheme could query exactly and achieve semantic content retrieval to some extent, which has made an improvement in efficiency compared with other type of query schemes.

Key words: video retrieval, global motion segmentation, semantic annotation, scenery videos