混合光子映射與光柵式渲染繪圖架構設計

Abstract

在本研究中，我們將提出利用光子映射與光柵式渲染混合的繪圖架構設計，藉由物件材質與陰影區域偵測以及三角形的光子密度資訊來建立混合渲染遮罩，利用該遮罩來決定該區域所要使用的渲染方式為光柵式渲染還是光子映射。利用混合渲染的方式在Cornell-box場景中，可得到比光子映射少61%的渲染時間，以及比光柵式渲染好的影像品質。同時也實做出光線走訪單元與三角形相交測試單元，利用硬體加速運算，來達到混合渲染的加速，硬體單元使用TSMC 90nm製程實現，與軟體在Intel® Core™ i5-4570 CPU上使用Microsoft Visual Studio 2013執行相比，當硬體單元在166 MHz下執行可加速約90倍。

In this thesis, we propose the hybrid rendering architecture for photon mapping and rasterization. The hybrid rendering mask for our hybrid rendering will be created by the object material, shadow area detection, and triangle photon density information. Then, the rendering method either uses rasterization or use photon mapping in this area will be determined by this mask. The hybrid rendering can provide about 61% lower rendering time than photon mapping, and better image quality than rasterization in Cornell-box. Furthermore, we implement the hardware units for traversal and intersection (T&I), and use the hardware unit to accelerate the operations. The hardware unit is implemented in TSMC 90nm process. The speedup with hardware unit operated at 166 MHz is about 90 times of the time executed on the software in Microsoft Visual Studio 2013 with Intel® Core™ i5-4570 CPU.