使用光線追蹤在OptiX框架下之蒙地卡羅路徑追蹤演算法分析比較

Abstract

傳統電腦圖學使用光柵化來繪製3D物件，但該方法對於擁有反射與折射等場景物件中，無法有較好的真實畫面。因此我們要使用基於物理性質計算的光線追蹤技術。在現實世界中，計算的資訊量是趨近於無限的，因此我們只能透過採樣來計算，即便如此運算的需求還是非常的高。 近年來計算機圖學領域隨著硬體設備的提升，即時的光線追蹤運算從二十年前的理論逐漸轉向可以實際達成。Nvidia推出的RTX系列圖形處理器與OptiX光線追蹤框架，提供使用者研究平行化部屬光線追蹤理論，本文中將使用OptiX 6.5 框架開發光線追蹤程式，研究不同場景與演算法之差異。主要探討演算法有原本圖學使用方法光柵化 (Rasterization) 與光線追蹤方法蒙地卡羅路徑追蹤(Monte Carlo Path Tracing)，雙向路徑追蹤(Bidirectional Path Tracing)，混合光線追蹤(Hybrid Ray Tracing)，比較各自演算法於不同圖形處理器和場景所產生渲染效果與速度。Traditional computer graphics uses rasterization to draw 3D objects, but this method does not have a good rendering for scene objects with reflection and refraction. Therefore, we need to use ray tracing technology based on the calculation of physical properties. In the real world, the amount of information calculated is close to infinite, so we can only calculate by sampling, even if the demand for such calculations is still very high. In recent years, with the improvement of hardware devices in computer graphics, real-time ray tracing operations have gradually become practical from the theory of 20 years ago. Nvidia's RTX series of graphics processors and OptiX ray tracing framework provide a simple, recursive, and flexible pipeline for accelerating ray tracing algorithms. In this paper, we will use the OptiX 6.5 framework to develop ray tracing programs to compare the differences between different scenes and algorithms. we discuss ray tracing algorithms Monte Carlo Path Tracing, Bidirectional Path Tracing, Hybrid Ray Tracing, and compare the effect and speed of each algorithm on different graphics processing unit.