We present an integrated framework on personalizedsports video customization, which addresses three research issues:semantic video annotation, personalized video retrieval and summarization,and system adaptation. Sports video annotation serves as the foundationof the video customization system. To acquire detailed descriptionof video content, external web text is adopted to align with therelated sports video according to their semantic correspondence. Basedon the derived semantic annotation, a user-participant multiconstraint0/1 Knapsack model is designed to model the personalized video customization,which can unify both video retrieval and summarization withdifferent fusion parameters. As a measure to make the system adaptive tothe particular user, a social network based system adaptation algorithmis proposed to learn latent user preference implicitly. Both quantitativeand qualitative experiments conducted on twelve broadcast basketballand football videos validate the effectiveness of the proposed method.
展开▼
