Organization of video data based on semantic representation by conceptual clustering

• Authors: Masayo Kaji;Kuniaki Uehara
• Venue: Information organization and databases
• Year: 2000

Video database systems should support easy and dynamic description process of video data, to manage them efficiently, and to provide content-based access to users. However, video content description process remains to be done manually, which means high cost and time consuming. Furthermore, video content description will always depend on user's interpretations. This paper proposes a dynamic method, in which the consumption time to describe stories is reduced and the content of the story remains the same regardless of different users' interpretations. We will discuss the organization of video databases to make user's queries more efficient. Our framework of representing of video data is based on descriptions of typical scenes organized in a hierarchical structure in which the same description can be reached by different paths. Furthermore, we can support an efficient retrieval by using the hierarchical structure. It is possible to maintain the consistency in the hierarchical structure even though the structure is dynamically changed by the addition of new descriptions.