中国科学院软件研究所机构知识库
Advanced  
ISCAS OpenIR  > 综合信息系统技术国家级重点实验室  > 学位论文
Subject: 计算机应用::计算机图形学
Title:
真实感绘制中自适应采样与重构研究
Author: 刘晓丹
Issued Date: 2014-05-22
Supervisor: 郑昌文
Major: 计算机应用技术
Degree Grantor: 中国科学院研究生院
Place of Degree Grantor: 北京
Degree Level: 博士
Keyword: 自适应采样 ; 各向异性重构 ; KD树 ; 并行采样 ; Contourlet小波 ; 光子映射 ; 回归分析
Abstract:

真实感图形绘制是计算机图形学的重要研究方向,其目标在于使用计算机精确生成逼近真实相机拍摄的图片,可以广泛应用于游戏电影、计算机辅助设计、模拟仿真和医学成像等领域。因为现有的真实感图形绘制方法在图像质量、绘制代价以及支持效果等方面还不能满足不断发展的应用需求,所以针对绘制空间、绘制消耗和绘制效果进行研究,提高算法的绘制速度和绘制质量具有重要的理论意义和实用价值。

本论文在全面分析和调研现有自适应采样与重构算法的基础上,重点研究了基于光线追踪和光子映射的方法,发现目前真实感绘制算法主要存在以下问题:1)多维自适应绘制算法存在维度灾难问题;2)基于频域的绘制方法存在采样点分布走样和各向同性重构问题;3)光子映射方法缺少自适应采样与各向异性重构方法;4)难以对采样点进行复用并绘制平滑的真实感图像。针对上述问题,本论文取得了以下创新性成果。

一、针对目前大多数多维绘制方法存在维度灾难的问题,提出了基于KD树的并行多维自适应采样方法。该方法利用KD树对绘制空间进行划分,提取并保存空间特征,给每个子空间分配采样点,在并行采样的同时保证采样点整体的自适应性,使绘制速度提升46倍;同时为了提高多维绘制中采样点分布的质量,利用KD树的结构在多维空间投放符合泊松盘分布的各向异性采样点,减少了绘制噪声和边界走样。

二、针对频域绘制方法中采样点分布走样和各向同性重构等问题,提出一种基于Contourlet小波的自适应采样和多尺度各向异性重构方法。该方法通过迭代分析绘制空间在各个尺度下的特征信息,避免了传统采样方法存在采样点分布走样的问题;同时利用方向滤波器组提取各向异性信息,针对每个像素构建多尺度各向异性滤波器,有效避免了绘制边界走样。

三、针对光子映射收敛速度慢以及缺少各向异性重构的问题,提出了基于光子映射的自适应采样与重构方法。该方法在光子采样过程中,通过划分空间,评估绘制空间中光照特征分布,构建累积分布函数,自适应地选择光子传播的路径,使得光子可以更多地投放到场景中的高频区域,提高了光子映射的收敛速度,可以应用于传统光子映射和渐进式光子映射;在重构过程中,利用场景中保存的光子,计算局部的各向异性张量,使用一种各向异性光子重构算法重构图像,减缓了生成图像中的走样和模糊现象。

四、针对难以复用采样点和绘制结果不平滑的问题,提出了基于回归分析的自适应绘制方法。首先,利用聚类算法组织绘制空间,通过最小二乘法构建多项式方程表示绘制空间中的光照变化信息,积分光照方程重构图像,该方法生成图像效果平滑;其次,利用人工神经网络构建全局光照模型,通过采样中得到的间接光照训练人工神经网络,重构时复用采样信息,提高绘制速度。

 

 

English Abstract:

Photorealistic rendering is an important research direction in Computer Graphics, with the goal of generating the real camera pictures. It has been widely used in entertainment, computer-aided design , simulation, medical imaging and other fields. But the existing photorealistic rendering methods cannot meet the improving requirements such as image quality, time consumption and rendering effect. Therefore, it has important theoretical and practical value to research adaptive sampling and reconstruction in photorealistic rendering.

In this thesis, based on a comprehensive analysis and research of the existing adaptive sampling and reconstruction technique, the following problems of current photorealistic rendering algorithms have been found: 1) multidimensional adaptive sampling has dimensionality curse problem, 2) adaptive wavelet rendering method has sample distribution aliasing and isotropic reconstruction issues, 3) photon mapping method converges slowly and has few anisotropic reconstruction methods, 4) it is difficult to reuse the samples and generate smooth photorealistic images. To solve these problems, this thesis makes the following contributions.

First, to solve the problem of dimensionality curse, a parallel KD-tree based multidimensional adaptive sampling method is proposed. At the beginning, the rendering space is organized and analyzed by KD-tree. Then, each subspace is assigned a certain sample budget which guarantees that the sample distribution of the whole space is adaptive. To improve the efficiency of samples, a multidimensional anisotropic Poisson disk sampling method is also proposed, which generates anisotropic blue noise samples. Compared with the previous methods, our method accelerates the rendering speed, reduces the noise and aliasing.

Second, to improve the aliasing of the adaptive wavelet rendering method, an adaptive sampling and reconstruction method using multi-scale and directional analysis is proposed. Our method bases on Contourlet transform. In sampling stage, the rendering space is analyzed by Laplacian Pyramid in multi-scale. In reconstruction stage, directional filter bank is used to compute the anisotropic information and build the per-pixel anisotropic filter to synthesize the final image. Compared with the previous image based methods, our method gives better results.

Third, to accelerate the convergence rate of photon mapping and improve the image quality, an adaptive importance photon shooting technique and an anisotropic progressive photon mapping method are given. In photon shooting stage, the rendering space is organized and analyzed. Then, the reflect direction of photon is adaptively chosen by the space feature, which uses a cumulative distribution function to locate more photons on the high frequency area. In reconstruction stage, an anisotropic tensor computed by the photons in the scene is used to reconstruct the final image.

Fourth, to reuse the sample and generate smooth result, a special reconstruction methods using regression analysis are proposed. Based on cluster sampling, least squares technique is used to model the light contribution in each clusters and generate smooth images by integrating the model functions. The artificial neural network is also employed in our method to train and reuse the indirect light. It can generate global illumination images in high speed. 

Language: 中文
Content Type: 学位论文
URI: http://ir.iscas.ac.cn/handle/311060/16472
Appears in Collections:综合信息系统技术国家级重点实验室 _学位论文

Files in This Item:
File Name/ File Size Content Type Version Access License
真实感绘制中自适应采样与重构的研究-刘晓丹.pdf(7027KB)----限制开放 联系获取全文

Recommended Citation:
刘晓丹. 真实感绘制中自适应采样与重构研究[D]. 北京. 中国科学院研究生院. 2014-05-22.
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[刘晓丹]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[刘晓丹]‘s Articles
Related Copyright Policies
Null
Social Bookmarking
Add to CiteULike Add to Connotea Add to Del.icio.us Add to Digg Add to Reddit
所有评论 (0)
暂无评论
 
评注功能仅针对注册用户开放,请您登录
您对该条目有什么异议,请填写以下表单,管理员会尽快联系您。
内 容:
Email:  *
单位:
验证码:   刷新
您在IR的使用过程中有什么好的想法或者建议可以反馈给我们。
标 题:
 *
内 容:
Email:  *
验证码:   刷新

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

 

 

Valid XHTML 1.0!
Copyright © 2007-2019  中国科学院软件研究所 - Feedback
Powered by CSpace