隨機光柵化在裁切空間下採樣點剔除技術

Abstract

在渲染真實的相機影像時，兩種現象是很常見的：動態模糊(motion blur)與景深模糊(defocus blur)。我們提出了一個針對動態與景深模糊在隨機光柵化技術下的裁切空間(clip-space)採樣點剔除技術。這個二階段的測試利用裁切空間下的資訊去降低我們需要做的覆蓋測試採樣點數量，包含鏡頭座標與時間座標下的所有採樣點。 首先我們做一個簡略的測試取得保守的鏡頭座標範圍值，並去除在此範圍之外的採樣點。在第二階段的測試時我們在裁切空間下針對每個三角形頂點找出xyuvt空間的相似三角形的關係，藉由此三角方程式去剔除不屬於此範圍內的採樣點。本篇論文提供了在即時運算的隨機光柵化中的簡單採樣點剔除方法，並且在少量的運算下可達到良好的採樣點測試效率(sample test efficiency)。

To render realistic camera images, two effects are common : motion blur and defocus blur. We present a novel clip space culling test of stochastic rasterization of motion and defocus blur. This 2-stage test use the clip space information to reduce the samples needed to be coverage tested over camera lens domain (uv) and time domain (t). First we do a rough test to get a conservative range of the camera lens uv bound, and cull the samples outside this bound. Then the second test finds a similar triangular equation for each triangle vertex in xyuvt space. Based on this equation, we cull the rest of samples outside. We present a simple method for the real-time stochastic rasterizer, and achieve a good sample test efficiency with low computation cost.