Real-Time Rendering Of Large-Scale Complex Scene

Real-time rendering of large-scale complex scene has been used widely in many area, either in online interactive entertainment,GIS, or Military. With the development of information society, many industry area need this rendering technique and meanwhile the requirement of image quality and rendering speed has become more and more strict. This is why this research area is hot.The biggest problem of this technique is how to solve the confliction between large data and limited computer resource. Large data mean the data is huge and these data usually cannot be loaded to memory at one time. So we need a strategy to do paging between system memory and disk. When data is loaded in, how to organize and use them to satisfy the real-time requirement is very important. Besides, Gap, lighting, texture are also need to be solved. In the previous work, people usually build a complex data structure such as BSP tree and k-d tree to hold the data. But these algorithm cost lots of CPU calculation and make CPU in heavy work. Finally, what this paper would solve is the decreasing the CPU load.This paper's research work mainly focuses at transforming CPU calculation to GPU and the algorithm is based on the previous Geo Clipmap algorithm. This paper makes some optimization on Geo Clipmap and has finally built a real-time rendering software.The innovation and contribution of this research work is making some optimization on previous Geo Clipmap algorithm and decreasing CPU load by transforming most CPU calculation to GPU.This paper's finally experiment also shows that this GPU algorithm does decrease the CPU load.In the following chapters, this paper would first introduce some previous related work, then introduce Geo Clipmap and its GPU implementation. Finally some experiment result would be given to prove this paper's research point.