| In real-time rendering of 3D scene, there exists a contradiction between the speed and the quality. When using advanced lighting method such as photon-mapping, it will cost a lot in both time and memory, and these advanced methods can't be used in virtual reality system. This thesis use radiositymap pre-compute the illumination on the surface of entities, and save these illuminations in textures called radiositymap, and map radiositymaps just like common textures while rendering. This can save a lot of computing time, and the difficulty change from fast ray tracing to how to map entities from 3D to 2D,so called "parameterization". Because it's a pre-computing problem, the radiositymap we used can associate with photon-mapping ,which is a very advanced lighting method nowadays, compute the illumination offline. When using radiositymap, there'll be a huge amount of textures, so we use texture packing and texture compressing to reduce the cost of memory and GPU.This thesis put radiositymapping, texture compressing, photon-mapping system and real-time rendering system together, constructed a photon-mapping based radiositymap computing system.. The photon-mapping based radiositymap computing system consists of parameterization subsystem, photon-mapping subsystem, and texture compressing subsystem. First this thesis introduced a variety of parameterization methods and put forward a normal-primary based parameterization method, which is extensible. Then it introduced the interface of parameterization subsystem and photon-mapping subsystem. The interface is clear and the photon-mapping system can be changed to other engines such as radiosity ,the only request is that the engine has the same interface. Then this thesis introduced texture compressing subsystem, put forward packing based texture compressing, can gain a higher compressing ratio than S3TC. At last this thesis gave 3 test results of different scenes. We can see that this system meet the demand of reality and real-time... |
PDF Full Text Request