容積陰影的時間適應性取樣演算法

Abstract

繪製場景時，若能考慮空氣中的介質對光的影響，則可以大大增加真實性，達到像是光束(light beam)或是神光(god ray)的效果，但是也因此必須考慮到光與空間中每個介質的交互作用，例如散射和吸收的影響，導致計算量大大的增加，而且若空氣中的介質並不唯一，也就是不同的介質在散射和吸收的特性不一樣，再加入此考量的話會更加的複雜，所幸只考慮單一介質的結果已非常逼真，故此論文以單一介質為基本假設。本論文的動機是考慮到取樣空間中既有的三個維度之外，還可以考慮時間維度下的取樣，針對動態場景，考慮動的物體所產生的容積陰影所影響到的像素範圍來做更新，而沒被影響到的區域使用前一畫面的資訊，如此一來，可以減少計算量，達到重複使用而得到加速的效果，並且維持一樣的品質。

Scattering in participating media generates volumetric lighting effects known as crepuscular or god rays. Such effects greatly enhance the realism in virtual scenes. However, they are inherently costly as scattering occurs at every point in sampling space and thus it needs costly integration of the light scattered towards the observer. This is often done using ray marching which is too expensive for each pixel on the screen for interactive applications. We propose a method to reduce samples in time domain, which means that we can reuse the information from the previous frame. In dynamic scenes, to decide pixels that can not be reused, we create shadow volumes of moving objects and rasterize them. Using a stencil buffer to maintain the information of screen pixels needs to recompute or just use the old information. We show that our method is simple to implement and yields the same quality but with a specific speedup.