Semantic video indexing using MPEG motion vectors

With the diffusion of large video databases and "electronic program guides", the problem of semantic video indexing is of great interest. In literature we can find many video indexing algorithms, based on various types of low-level features, but the problem of semantic video indexing is less studied and surely it is a great challenging one. In this paper we present a particular semantic video indexing algorithm based on motion information extracted from MPEG compressed bit-stream. This algorithm is an example of solution to the problem of finding a semantic event (scoring of a goal) in case of specific categories of TV programmes.