Image Quality and Performance Comparison between GPU Ray Tracing and Rasterization

• Main Authors: Chin Chieh Chuang, 莊沁傑
• Other Authors: Chun Fa Chang
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/84373853428142458731

碩士 === 國立臺灣師範大學 === 資訊教育學系 === 102 === With faster graphics processors and GPGPU languages, we have the ability to compute more complex data. However, we found few people want to simulate ray tracing on the graphics processor which can render high quality pictures. The developers think the cost of generating one frame is expensive by tracing rays. Besides, they are not willing to learn new GPGPU languages such as OpenCL and CUDA, because of the lack of powerful and convenient engines. In order to achieve that purpose, we offer two methods: One Pass Rendering Algorithm and Hybrid Rendering Algorithm to implement ray tracing on Shader and also implement the Phong lighting effect and the environment reflection effect. We compare the image quality and performance between rasterization and our Shader based ray tracing by different kinds of scenes and we also quantify the results of this comparison. We show the difference between GPU ray tracing and rasterization and we explain the pros and cons of them. The performance of rasterization is better than ray tracing, but we do not need to be worry about it. Because we do not get the larger performance gaps between GPU ray tracing and rasterization by increasing effects. On the contrary, we can gain higher image quality by using ray tracing technology. According to this experiment, we successfully prove the possibility of ray tracing simulation technology on GPU.