Global Illumination of large Scenes using Illumination-driven Geometry Adaptation

蒙特卡罗方法是当今物理校正全局照明模拟的标准方法。尽管场景大小永久增长，但仍可以计算这样的模拟。即使是核外场景也是可能的，但是尽管使用了大量的并行硬件，但是交换几何数据的时间消耗很大，因此，该项目的目标是对运行在主内存有限的单机上的这种核外场景进行光照驱动的几何简化。给定摄影机、材质和光源的当前设置，生成的照明模拟的质量应该会变得最优。因此，所需的几何图形减少过程应集成到照明模拟中。简化网格中区域的标准应该是照明中的可见变化，这将导致网格的细节级别适应全局照明。这尤其对应于在相机图像中不直接可见的区域，但间接地对可见照明有贡献。该项目的目标是在减少主内存的情况下，确定物理校正的全局照明模拟的最佳质量。本项目建议书是DFG项目《核外全球照明》的延续。

Monte carlo methods are todays standard for physically correct global illumination simulations. Such a simulation can be computed despite the permanently growing scene size. Even out-of-core scenes are possible, but with a large time consumption for swapping geometry data in and out, although a large amount of parallel hardware is used.Therefore, this project aims for an illumination-driven geometry simplification of such an out-of-core scene which runs on a single machine with limited main memory. Given the current settings for camera, materials and light sources, the quality of the resulting lighting simulation should become optimal. The required geometry reduction process should therefore be integrated in the lighting simulation. The criterion for simplification of a region in the mesh should be the visible change in the illumination, which then leads to an adaptation of the level of detail of the mesh to the global illumination. This corresponds especially to regions which are not directly visible in the camera image, but contribute to the visible illumination indirectly. The goal of the project is to determine the best quality of a physically correct global illumination simulation, given a reduced amount of main memory. This project proposal is the continuation of the DFG project "Out-of-Core Global Illumination".