Dynamic Video Content Decomposition: An Efficient Scheme for Personalized Video Content Navigation over Τhe Web

In this paper, an adaptable hierarchical video summarization scheme is proposed which dynamically organizes video files to fit the current user’s information needs and preferences. The architecture non-linearly (non-sequentially) decomposes a video sequence into content resolution levels of representative shots/frames based on a tree structure scheme. However, instead of a static tree construction, video decomposition is dynamically performed to fit the current user’s information needs and preferences. In particular, each time the user selects a tree node for further navigation, the node is considered as relevant. On the contrary, irrelevant nodes are marked as those nodes in which the user backtracks to previous levels. In the sequel, an on-line learning strategy is adopted for estimating the user’s profile according to the set of relevant/irrelevant selected nodes. Then, the decomposition tree is updating to meet user’s needs. Experimental results and comparisons with other conventional static schemes indicate the outperformance of the proposed architecture compared to other approaches. 1