A Global-Motion Analysis Method via Rough-Set-Based Video Pre-classification

摘要

Motion information represents semantic conception in video to a certain extent. In this paper, according to coding characteristics of MPEG, a global-motion analysis method via rough-set-based video pre-classification is proposed. First, abnormal data in MPEG stream are removed. Then, condition attributes are extracted and samples are classified with rough set to obtain global-motion frames. Finally, their motion models are built up. So the method can overcome disturbance of local motion and promote veracity of estimations for six-parameter global motion model. Experiments show that it can veraciously distinguish global and non-global motions.