Spatio-temporal structure analysis on motion imagery using syntactic approach

Video pattern analysis is becoming increasingly important because of the explosive increase in the production of multimedia data. Traditional statistical pattern analysis approaches, however, are not sufficient for analyzing this type of imagery because semantically similar events may have significantly different statistical properties, or even they may range over different time intervals. In such scenarios, a syntactic approach provides a better solution. This paper describes our approach to building a prototype syntactic video analysis system. For this example we concentrate on basketball video sequences. Our experimental results are promising, suggesting that syntactic pattern analysis may be a vital tool in video analysis.