Please use this identifier to cite or link to this item:
標題: 螺旋管面之均等取樣技術
作者: 王宗銘 
關鍵字: 取樣;均等取樣;真實影像合成;蒙地卡羅法
出版社: 國立中興大學工學院;Airiti Press Inc.
Project: 興大工程學刊, Volume 13, Issue 3, Page(s) 171-180.

Efficient and stratified sampling for various geometries has been the subject of intensive research in computer graphics community. A popular allocation is for realistic image synthesis where geometries represented as different light sources are sampled before applying Monte Carlo methods to solve the global illuminatin problem. In this paper, we present a novel stratified sampling algorithm (SSA) for the helical canal surface to extend the coverage of the light sources. As our knowledge, no stratified sampling approaches for the helical canal surface have ever been addressed in the literature. Initially, we define the mathematical expression for a helical canal surface. We then present the SSA developed from the inverse function of the Integration on helical canal surface. Furthermore, we develop an acceleration technique, Faster SSA (FSSA), to meliorate the computational efficiency, significantly reducing the time required to generate stratified samples with negligible expenses. We have implemented our algorithm on an experimental system where the implementation is effortless. In this system, the front-end user can visualize samples being generated interactively to illustrate the visual effect of SSA. Furthermore, the FSSA expedites the sample generation process considerable with the expense of little sample deviation. With FSSA, 2000 samples can be general within 0.13 second, which is almost 678.9 times faster than employing conventional SSA. The sample deviation is negligible since the average deviation for canonical number is relatively small, being less than5.0�10-7 in average for 2000 samples. In conclusion, the proposed stratified sampling algorithm is novel, efficient and feasible for a helical canal surface.
ISSN: 1017-4397
Appears in Collections:工學院
第13卷 第3期

Show full item record

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.