Logo image
基於繪圖處理器的即時光線追蹤及全域照明技術之研究
Dissertation

基於繪圖處理器的即時光線追蹤及全域照明技術之研究

陳英傑
Doctor of Philosophy (PHD), 國立清華大學, 資訊工程學系
2011

Abstract

電腦圖學 光線追蹤 全域照明 繪圖晶片上的一般目的計算 Computer Graphics Ray Tracing Global Illumination GPGPU
In the past decade, graphics hardware has improved dramatically, especially on the capabilities of graphics processing units (GPUs). The most prominent improvements are realization of programmable shaders for real-time rendering and general purpose computation with graphics processing units (GPGPUs) for advance rendering. The exquisite geometry details and vivid surface shading are therefore produced in real time. Modern games usually present the geometry details with a huge number of small-area primitives to improve the quality of final videos. To render such scenes, inverse displacement mapping methods, one kind of ray-tracing-based method, are a better choice than the rasterization approaches (used in traditional real-time 3D graphics) which is designed for large-area primitives. However, existed inverse displacement mapping methods are limited to planar surface. To eliminate the constraint, I propose an arbitrary-surface-supported inverse displacement mapping method in 2. With the proposed algorithm, the silhouette and geometry detail are rendered correctly in real time. In computer graphics, “Real-time” global illumination is a big challenge. It still takes minutes to days to synthesize a single image, even though the newest graphics hardware (GPGPUs) speeds up the process by 2~100 times. Thus, not only computing power but also better algorithms are required to achieve real time. In 3, a spatio-temporal approach for the low-frequency indirect lighting is proposed. This method amortizes the cost over a sequence of image frames. As the result, the computing power requirement is dramatically reduced, and the indirect lighting effect is rendered in real-time.

Metrics

1 Record Views

Details

Logo image