Interactive system of real-world video based on self-organized data structure

With recent development of digital broadcasting and networks, the volume of image data to be handled has rapidly increased. To handle massive amounts of image data on a network, the authors have proposed "advanced database TV" (ADTV) in which image data on the network can be shared. This paper proposes a real-world video system which is a realization of ADTV, with examples. In this paper, a method of a self-organized data structure based on image objects is proposed, since the structural elements of images (e.g., scenes, cuts, and objects) are essential elements. The method can be realized by matching a realworld video of a city and its map. It is shown that a significant improvement of the matching accuracy can be achieved by obtaining three-dimensional information from the images.