Approach of MTV abstraction automatic extracting based on shot-group weighting

An approach based on shot-group weighting to generate static and dynamic MTV abstraction was proposed. The static abstraction was generated by obtaining the shot-group information in MTV, and the dynamic abstraction was generated by hierarchical analyzing. Experiment shows that static abstractions achieve 86.25% average recall with 86.25% average precision, and dynamic abstractions can reflect the main contents of the original videos with good intelligibility and appreciation quality when compression ratios are lower than 5.