Some Optimization And Acceleration Approaches For Ray Tracing System

Ray tracing based rendering technique is an important research domain in Computer Graphics. It plays an essential role in many applied fields such as digital entertainment, scientific visualization, computer generated films and animations, and virtual reality.In this paper, we first make a briefly introduction on the background and state-of-the-art of modern graphics rendering techniques, especially on hardware rasterization based rendering and ray tracing based rendering. Then we focus on discussing two main methods of photo-realistic rendering: ray tracing based and pre-computed radiance transfer based global illumination, with their specifications, implementation strategies, applying scope and limitations.Rendering methods based on ray tracing are closer to physical illumination model, therefore more accurate and tending to generate more delicate and subtle images, but they have many limitations, such as high computation cost, hard to apply to modern hardware. Since this kind of techniques need a huge number of samples from light source, geometry surface, material, and ray path in the scene which forms a mass of high dimensional data, their rendering speed are rather slower compared to modern hardware rendering, which makes them hard to apply to real-time applications.In this paper, we make some important optimization and accelerations for a photo-realistic ray tracing system, including optimizing KD-Tree acceleration structure, SIMD ray-intersect computing, optimizing photon mapping process, fast visibility computing for indirect lighting, generating better image film samples, taking better samples for Monte Carlo estimator, and multi-thread acceleration etc. By using these optimization approaches, we could obtain faster rendering speed and better rendering quality.