Bilkent University
Department of Computer Engineering


Stereoscopic Visualization of Urban Scenes


Turker Yilmaz
Ph.D Student
Computer Engineering
Bilkent University

In this presentation, we are going to describe a fast way of visualizing urban sceneries in stereo. Fast urban visualizations require fast intermediate solutions to the problems of urban scene visualization. These are navigation space determination, preprocessed occlusion culling and stereoscopic visualization. We start by giving some taxonomies on visualizing urban scenes and describe deficiencies related with the current state of the art work. Next, we describe our proposed framework along with the algorithms and contributions achieved. Visualization of urban scenery with a tight approximation of the visible part of the scene requires efficient occlusion culling algorithms and data structures. In this context, we describe an efficient tight conservative occlusion culling algorithm, which brings a solution to the partial occlusion problem. A conservative visibility culling algorithm accepts an object as completely visible even only a small portion of it is unoccluded. In our algorithm, we tried to avoid this by decomposing the objects into slices to achieve tight conservativeness. This sliced structure is queried for visibility at predefined grid locations in the scene. Since the slices of objects are checked for visibility instead of complete objects, only a very small portion of polygons of an object are sent to the graphics pipeline unnecessarily if a small portion of the object is unoccluded. For the stereoscopic visualization we will make use of the view cells created during preprocess and use transformations of the visibility lists, instead of processing them twice.


DATE: November7, 2005, Monday@ 16:40