In this paper we present a VRML97-X3D extension to describe pre-computed visibility relationships in the context of progressive transmission as well as real time visualization of massive 3D sceneries. The extension we propose can be used to describe cell-to-cell, cell-to-objects as well as hybrid visibility relationships using a generic cell representation. Thanks to these relationships it is possible to represent indoor and outdoor sceneries using the same VRML97-X3D representation. We also present some new mechanisms that allow to perform instance sharing of objects described into separate files. These mechanisms allow to minimize the size of the set of VRML97-X3D files that are used for the database description as well as the amount of main memory required on the client side to visualize the scenery. In addition to these new mechanisms we also explain how we manage the memory on the client side and how we perform data pre-fetching according to the viewpoint movements. Finally we present several tests performed on different kinds of sceneries such as city and architectural models.