Font Size: a A A

The local Z buffer algorithms in real time rendering of large complex scenes

Posted on:2005-11-22Degree:Ph.DType:Dissertation
University:The University of IowaCandidate:Han, XiaoxuFull Text:PDF
GTID:1458390008493390Subject:Computer Science
Abstract/Summary:
The large scene complexities from large complex scenes challenge the computer graphics hardware and real-time rendering algorithms. The real-time rendering algorithm design under such environments has been a core research topic in computer graphics since 1990s.; In this work, we present a real-time rendering algorithm LZB (local Z buffering) for large complex scenes. A naive local Z buffering algorithm, which works well for the near-view densely occluded scenes, is proposed firstly based on a rendering framework, The naive LZB uses Z buffering to render the near scene Snear and use ray casting to render the far scene Sfar by partitioning the input scene S into a disjoint union S = Snear ∪ Sfar. The naive LZB algorithm is refined by a series of techniques such that the local Z buffering algorithm is an output sensitive algorithm for rendering general large complex scenes. The approaches include lazy ray-casting, object oriented ray casting (0011), static and dynamic local occluder selection, a coherence based octree traverse algorithm and hybrid rendering methods.; The lazy ray casting gives hardware support in the ray casting by finding a potential visible set for each pixel and using Z buffering to render the final potential visible set (FPVS). The object oriented ray casting (OOR) removes empty/background pixels involved in the ray casting efficiently. Static and dynamic local occluder selection give priorities to the highly likely visible triangles rather than treat all triangles uniformly in the ray casting process such that the FPVS can be found quickly. A coherence based octree traverse algorithm is to cache the previous ray path and reuse it in the current ray traverse to exploit pixel coherence.; Compared to the other real-time rendering algorithms, the LZB is easy to implement, to parallelize. The LZB can achieve satisfied speed-up averagely compared to the Z buffering based rendering for large complex scenes. The performance of the LZB algorithm can "grow up" with the CPU and GPU technology since it is an integration of CPU and GPU based rendering approaches.
Keywords/Search Tags:Rendering, Large complex scenes, Algorithm, Local, Ray casting, LZB
Related items