A scalable approach to visualization of large virtual cities

Visualization of large urban complexes on the Web is a highly demanding task both from the networking and computational point of view. The whole three-dimensional model of a city is mostly too big to be downloaded in a reasonable time and to be stored in a memory of commonly used computers. This paper presents a scalable approach for subdividing an urban model into smaller parts. Dynamic loading and unloading data allows theoretically unlimited extension of a model with constant network and computation load. Further requirements specific to presentation of virtual cities are also discussed. Theoretical considerations are followed by a practical implementation utilizing the VRML language for modeling the historical city of Prague.