Exploitation of Temporal Dependencies of Descriptors to Extract Semantic Information

Nowadays the problem of semantic video indexing is of great interest due to the wide diffusion of large video databases. In this paper we present a semantic video indexing algorithm based on finite-state machines that exploits the temporal dependencies of low-level descriptors extracted from the MPEG compressed bit-stream. The performance of the proposed algorithm have been evaluated considering the semantic indexing of soccer games video sequences, and the simulation results have shown the effectiveness of the proposed algorithm.