Efficient Occlusion Culling and Non-Refractive Transparency Rendering for Interactive Computer Visualization

This dissertation, "Efficient Occlusion Culling and Non-refractive Transparency Rendering for Interactive Computer Visualization" by Chun-ho, Poon, 潘仲豪, was obtained from The University of Hong Kong (Pokfulam, Hong Kong) and is being sold pursuant to Creative Commons: Attribution 3.0 Hong Kong License. The content of this dissertation has not been altered in any way. We have altered the formatting in order to facilitate the ease of printing and reading of the dissertation. All rights not granted by the above license are retained by the author. Abstract: Abstract of thesis entitled Efficient Occlusion Culling and Non-Refractive Transparency Rendering for Interactive Computer Visualization submitted by POON Chun-Ho for the degree of Master of Philosophy at The University of Hong Kong in August 2000 Hidden surface removal and non-refractive transparency rendering are two fundamental problems for interactive visualization. Hidden surface removal helps to identify the hidden primitives, bypass their rendering processes, so as to reduce the workload of graphics engine and provide an interactive frame rate. Without the distortion from refraction, non- refractive transparency rendering gives more insight of the interior structure of geometric models. This research is aimed at studying these critical problems. Exact hidden surface removal is usually implemented at pixel level, such as z-buffering, in order to achieve the image precision. It is computationally power demanding. A conservative algorithm that finds a minimal superset of visible primitives in a short time is considered as the most suitable approach for interactive visualization. One significant technique that makes use of occlusion relation among the primitives of the model, and culls a significant amount of invisible primitives at different viewpoints quickly, is studied and devised to an occlusion culling algorithm. In this thesis, we present two algorithms to improve occlusion culling for a highly occluded virtual environment. The first is a novel method to select occluders with multiple criteria at pre-processing stage for static environment, using the idea of the minimum occluder set (MOS). The MOS of an occludee is the minimal set of primitives that occludes the occludee. The second is an efficient occlusion culling algorithm using the opacity map (OM) and sparse depth map (SDM), which are applied to the spatial hierarchy of the whole model at each frame at run-time. A common method for non-refraction transparency rendering is interpolated transparency. Interpolated transparency calculates the resultant intensity by linearly interpolating the corresponding intensities of two pixel fragments. The main disadvantage of interpolated transparency is that, when there are more than two pixel fragments, the graphics pipeline has to apply pair-wise interpolations in a sorted visibility order. However, finding visibility order at object space is a complex task, especially for a self-penetrating model. Another approach is screen door transparency. Screen door transparency generates a set of two dimensional masks that represents the different levels of opacity. The pixels of the masks have only two states, which determine whether they are visible or not. At higher opacity level, the corresponding mask contains less filled pixels. Though this technique is order invariant, the correctness of resultant opacities depends heavily on the size of mask and the distribution of pixels inside each mask. An improved screen door transparency rendering is presented, that applies a simple and fast mask generation algorithm and is feasible to be implemented at sub-pixel level. The algorithms mentioned above have been implemented and applied to different experimental tests and a learning toolkit for medical student called "Virtual Brain." A significant speedup and a high acc