Automatic Movie Trailer Generation Based on Semantic Video Patterns

Automatic video summarization has become an important field of research with the advances in digital audio and video analysis and much effort has been put into movie abstracting for large media databases. However, in the movie industry content summarization for advertising trailers has been perfected to a form of art. In this paper we introduce the approach of automatically generating entertaining Hollywood-like trailers based on a trailer grammar. The extraction of features from movies using state-of-the-art image and audio processing techniques builds the foundation for the selection of meaningful and usable material which is re-assembled according to defined grammar rules. We present a system for generating trailers for contemporary Hollywood action movies. User testing of our automatically produced trailers for this movie genre shows promising results that suggest further research in this field.