Abstract
In urban scene walkthrough, many researchers focused on how to compute CVS(Conservative Visible Set) more efficiently and quickly. However, a basic but important question about how to define viewcell properly is rarely mentioned. By exploring the spacial coherence, for example, many viewcells defined along a street may have similar visibility, we propose a new viewcell construction algorithm. Moreover, during remote walkthrough, server must compute EVS(Exact Visible Set) for clients as soon as possible and only transmit visible polygons to the clients. If viewcell switching occurs too frequently, server will not finish the EVS computation in time, and the clients will have incomplete images. In this paper, a new method is presented to construct viewcells based on similarity coherence. By using this method, the total viewcell number can be greatly reduced and the CVS difference between viewcells becomes more significant. Several examples are given to show that during remote walkthrough, the proposed method can reduce the number of viewcell switching and greatly improve the efficiency of server.