Visibility Preprocessing with Occluder Fusion for Urban Walkthroughs

Abstract

Abstract. This paper presents an algorithm for occlusion culling from regions of space. It is conservative and able to find all significant occlusion. It discretizes the scene into view cells, for which cell-to-object visibility is precomputed, making on-line overhead negligible. Unlike other precomputation methods for view cells, it is able to conservatively compute all forms of occluder interaction for an arbitrary number of occluders. We describe an application of this algorithm to urban environments. A walkthrough system running an 8 million polygon model of the city of Vienna on consumer-level hardware illustrates our results.

Conservative occlusion culling for urban visualization using a slice-wise data structure

“…In order to achieve conservative occlusion culling, we made use of the shrinking idea first proposed by Wonka et al [10]. Since in our models we do not limit model complexity, we developed a new shrinking method, which can be applied to any kind of scene object, including nonconvex ones.…”

“…If a point P1 is visible from the center of the view cell, then it should also be visible with respect to its shrunk version S, even if the user moves to the farthest distance available in the view cell, ". If it is defined that the inner movement distance d of the faces for the shrunk shape calculation is greater than or equal to ", the conservativeness is guaranteed and the point becomes visible with respect to the shrunk version S. The reader is referred to [10] for the proof of the other case: if a point is occluded with respect to the shrunk version S, then it is also occluded with respect to its original version O, within an " neighborhood. …”

“…Approximate occlusion-culling algorithms, such as [7], [8], [9], the visible primitives are rendered up to a specified threshold, i.e., some of them may not be sent to the graphics pipeline although they are visible. There are also approaches to occlusion culling that use parallel processing approaches, such as [10], [5], [11].…”

“…Visibility determination by traversing a scene hierarchy requires the quick selection of occluders; this is a difficult task [10], [13], [14], [15], [16], [17], [18].…”

Occlusion horizons for driving through urban scenery

“…This requires the frame buffer to be read for each view point, which is slow on some architectures and limits the accuracy of the algorithm to the resolution of the frame buffer. Later, this technique was extended to a cell-based method and used as a preprocessing step [10]. The upgraded algorithm organizes the information hierarchically, but the preprocessing still takes about 9 hours for 8 million polygons, where more than 50% of the time is spent reading the frame buffer and the extracted visibility information requires a good deal of storage space.…”