A Shading Reuse Method for Efficient Micropolygon Ray Tracing

晃晃

A Shading Reuse Method for Efficient Micropolygon Ray Tracing

Qiming Hou Kun Zhou

State Key Lab of CAD&CG, Zhejiang University

Abstract

We present a shading reuse method for micropolygon ray trac-

ing. Unlike previous shading reuse methods that require an ex-

plicit object-to-image space mapping for shading density estima-

tion or shading accuracy, our method performs shading density con-

trol and actual shading reuse in different spaces with uncorrelated

criterions. Specifically, we generate the shading points by shoot-

ing a user-controlled number of shading rays from the image space,

while the evaluated shading values are assigned to antialiasing sam-

ples through object-space nearest neighbor searches. Shading sam-

ples are generated in separate layers corresponding to first bounce

ray paths to reduce spurious reuse from very different ray paths.

This method eliminates the necessity of an explicit object-to-image

space mapping, enabling the elegant handling of ray tracing effects

such as reflection and refraction. The overhead of our shading reuse

operations is minimized by a highly parallel implementation on the

GPU. Compared to the state-of-the-art micropolygon ray tracing al-

gorithm, our method is able to reduce the required shading evalua-

tions by an order of magnitude and achieve significant performance

gains.

Keywords: micropolygon, GPU, Reyes, ray tracing



1 Introduction

Shading is typically the performance bottleneck in cinematic-

quality rendering, which is often based on the Reyes architecture

and uses micropolygons to represent high order surfaces or highly

detailed objects [Cook et al. 1987]. In order to reduce shading costs,

state-of-the-art micropolygon renderers (e.g., Pixar’s RenderMan)

perform shading computation on micropolygon vertices, and reuse

the shading values to evaluate the color of each visibility sample (or

antialiasing sample) and composite the final image. Such a shad-

ing reuse strategy enables a shading rate significantly lower than

the visibility sampling rate, which is vital for efficient high-quality

rendering where extremely high supersampling of visibility is nec-

essary, especially when rendering defocus and motion blur.

Existing shading reuse methods for micropolygon rendering are

primarily designed for rasterization based pipelines. Ray trac-

ing effects such as reflection and refraction are typically consid-

ered as a part of shading in such methods. Consequently, all re-

flected/refracted samples have to be shaded, incurring significant

overhead. As ray tracing achieves more significance in modern

high-quality rendering [Parker et al. 2010], this may become a ma-

jor obstacle in future applications.







In this paper, we introduce a simple but effective method to reuse

shading evaluations for efficient micropolygon ray tracing. Com-

pared to the state-of-the-art micropolygon ray tracing algorithm,

our method is able to reduce the required shading evaluations by

an order of magnitude and achieve significant performance gains.



1.1 Related Work

Extensive research has been done on micropolygon rendering and

ray tracing.

Researchers have explored efficient parallel implementations of mi-

cropolygon rendering on GPUs [Wexler et al. 2005; Patney and

Owens 2008; Zhou et al. 2009; Hou et al. 2010]. In particular,

Hou et al. [2010] introduced a GPU-based micropolygon ray trac-

ing algorithm. They demonstrated that for high-quality defocus and

motion blur ray tracing can greatly outperform rasterization meth-

ods. In their method, ray tracing is only used for visibility sampling

and shading is still performed on micropolygon vertices. Another

branch of research also seeks to accelerate micropolygon rendering

using GPUs [Fisher et al. 2009; Fatahalian et al. 2009; Fatahalian

et al. 2010; Ragan-Kelley et al. 2011; Burns et al. 2010]. The key

difference between our work and theirs is that while they propose

new GPU architecture designs that support real-time micropolygon

rasterization, we aim to accelerate high-quality, off-line ray tracing

using software approaches on current GPU hardware.















全文请下载附件：

菜刀吻电线

不错啊　经典

晃晃

呵呵,真得不错哦!!

奇

很经典，很实用，学习了!

晃晃

很经典，很实用，学习了!

奇

其实楼主所说的这些，俺支很少用！