Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/6250
Title: | 基於 SURE 之最佳化的自適應採樣與重建技術 SURE-based Optimization for Adaptive Sampling and Reconstruction |
Authors: | Tzu-Mao Li 李子懋 |
Advisor: | 莊永裕(Yung-Yu Chuang) |
Keyword: | 採樣,重建,光線追蹤,濾波器, sampling,reconstruction,ray tracing,cross bilateral filter, |
Publication Year : | 2013 |
Degree: | 碩士 |
Abstract: | 本論文提出了一個利用Stein's Unbiased Risk Estimator (SURE)的自適性採樣與重建演算法,以應用在蒙地卡羅影像渲染上。SURE是一個在統計上對均方差的無偏估計,利用SURE,我們可以估計任意一種重建濾波器所造成的誤差,使得我們可以使用更有效率的濾波器,例如使用輔助特徵圖的交叉雙邊濾波器(cross bilateral filter)或是非局域均質濾波器(non-local means filter),而非過去的方法所限制的對稱濾波器;此外,SURE也可以讓我們對估計出較高錯誤的區域使用更多樣本。由此我們可以建立一個最佳化SURE的自適性採樣與重建系統。本論文另外也提出一個對記憶體較友善的方法來減少景深與物體或視點移動造成的雜訊對用來輔助濾波的特徵圖的干擾。實驗顯示本論文提出的方法相較於之前的方法在同樣的時間內可產生更清晰,但較少雜訊的圖片。 This thesis presents a method applying Stein's Unbiased Risk Estimator (SURE) to adaptive sampling and reconstruction to reduce noise in Monte Carlo rendering. SURE is a general unbiased estimator for mean squared error (MSE) in statistics. With SURE, we are able to estimate error for an arbitrary reconstruction kernel, enabling us to use more effective kernels, such as cross bilateral filters utilizing auxiliary feature buffers or non-local means filter, rather than being restricted to the symmetric ones used in previous work. It also allows us to allocate more samples to areas with higher estimated MSE. Adaptive sampling and reconstruction can therefore be processed within an optimization framework. We also propose an efficient and memory-friendly approach to reduce the impact of noisy geometry features where there is depth of field or motion blur. Experiments show that our method produces images with less noise and crisper details than previous methods. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/6250 |
Fulltext Rights: | 同意授權(全球公開) |
Appears in Collections: | 資訊工程學系 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-102-1.pdf | 57.27 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.