Design of video scene databases with mapping to virtual CG space

This paper presents a new approach of modeling video scenes by adding meaning to the shape of real world objects and simplifying it. In order to store shapes and/or motions of objects that appear in real world scenes into the database, expressing the shape and its meaning and handling them efficiently are indispensable. For that purpose, we introduce a method for constructing a scene database based on an approximate polygon model that reflects the individual shape and meanings of an object, called shape of individual model. It is particularly suitable for modeling the human head. The scene database is constructed using the Shapes of Individual Models and their 3-dimensional motion by mapping then into virtual CG space. And this paper also discusses the evaluation of the created Shape of individual Model in an application of the scene database