以雲隙光為例之體積渲染效能研究

Abstract

一般渲染為了計算方便，經常假設場景為真空，然而真實世界的大氣，散布著許多微小粒子，為非真空狀態。遊戲或電影場景中，常見體積渲染的蹤跡，像是光暈、體積霧（volumetric fog）、雲隙光（godray）等效果，可為畫面增添朦朧迷霧感，這是真空場景中所無法實現的。體積渲染由於額外考慮光和參與介質的能量散射、碰撞，使得計算時間成本大幅增加，此為體積渲染的一大難題。隨著圖形處理器的進步，以及Nvidia推出的OptiX平行運算框架，體積渲染已有不錯的效能提升。本論文借鑒即時渲染常用以簡化的單一散射概念，結合OptiX光線追蹤引擎，針對多重散射進行改良，期望能在更接近即時的時間內實現god rays效果。結果顯示，在維持畫面真實無偏的同時，可達到更快的收斂速度。

In general rendering, for the convenience of calculation, it is often assumed that the unoccupied space of the scene is vacuum. However, the atmosphere of the real world is scattered with many tiny particles, which is not vacuum. We can see that volume rendering are common in game or movie scenes, such as halo, volumetric fog, god ray and other effects. Volume rendering can add a hazy and foggy feeling to the picture, which cannot be achieved in a vacuum scene.Volume rendering takes the energy scattering and collision of light and participating media into account, which greatly increases the calculation time cost, making it a major problem for volume rendering. With the advancement of graphics processors and the OptiX parallel computing framework introduced by Nvidia, volume rendering has improved its performance. This paper draws on the single scattering concept commonly used in real-time rendering to simplify volume rendering, combined with the OptiX ray tracing engine to improve the multiple scattering, and hope to achieve the god rays effect in a efficirnt time. The results show that while maintaining the true and unbiased picture, a faster convergence rate can be achieved.