中国科学院软件研究所机构知识库
Advanced  
ISCAS OpenIR  > 中科院软件所  > 中科院软件所
题名:
基于图象的快速绘制技术的研究
作者: 郑新
答辩日期: 2001
专业: 计算机应用技术
授予单位: 中国科学院软件研究所
授予地点: 中国科学院软件研究所
学位: 博士
关键词: 基于图象的绘制 ; 三维变换 ; 正向映射 ; 逆向映射 ; 极线 ; 极点 ; 层次纹理映射 ; 纹理映射 ; 绘制
摘要: 基于图象的绘制(Image-based Rendering)技术是目前计算机图形学的研究热点之一。与传统基于几何造型的绘制技术相比较,它具有造型简单、成象复杂度与场景复杂度无关、真实感强等许多优点。因此自九十年代初被提出以后,该技术引起了学术界和工业界的广泛关注,在短短几年内得到了迅速发展,逐渐成为计算机图形学中一个具有强大生命力和应用前景的重要分支,被广泛应用于虚拟现实、动画等许多领域。IBR技术是一个比较新的课题,研究起步较晚,仍然有许多问题如空洞、多幅图象合成的效率等需要解决。本文对IBR技术中一些理论和应用问题进行了深入分析和探讨,提出了一些基于深度图象的快速绘制算法。其研究工作以及创新思想包括以下几个部分:1)基于深度预测的多幅参考图象合成算法。在论文第二章中,本文提出了一种多幅图象合成算法。该算法首先进行主参考图象位移值的正向映射,以获取目标图象所对应的位移图;然后通过一个深度预测算法,对尚未完整的位移图作空洞填补:最后由逆向映射过程来生成最终目标图象。与已有的三维变换方法相比较,该方法不但成功地填补了由于投影区域扩张而产生的第一类空洞,而且成功地填补了由于空间深度非连续物体相互遮挡而产生的第二类空洞,从而方便地实现了虚拟环境另的漫游;基于物体表面深度的连续性,本文提出了一个位移预测方法—此方法可以从单幅参考图象获得逆映射过程中所需要的目标图象的位移信息,从而大大提高了处法的效率;与通常的正向映射算法相比,此算法克服了多幅参考图象所带来的计算量成倍增长等问题,而且误差较小。2)基于极线几何的快速逆映射算法。利用参考图象的边界信息与隐含的遮挡关系,以及极线几何的性质,本文第三章提出了一个基于极线几何的快速逆映射算法,从多幅参考图象精确合成当前视点目标图象。算法首先对参考图象作简单预处理,由深度撮参考图象的边界信息。然后从参考图象中选出一幅作为主参考图象,逐次分段处理目标极线,以生成目标图象。最后从其它参考图象搜索对应象素以填补目标图象中的空洞。利用极线几何的性质和参考图象的边界信息,该方法加速了逆映射过程中对应匹配点的搜索。由于映射过程中,只需处理参考图象的边界点对,即可得到对应目标极线上所有点的深度信息,因此大大提高了成象速度和精度。与已有的逆映射算法相比,新方法便于处理深度非连续的图象。3)层次纹理映射。在第四章中,本文提出了层次纹理映射算法。该算法将深度图象按深度分层存放,然后利用纹理映射的硬件支持,用通常的纹理映射方法将各LT图象由后往前依次映射到成象面上就能生成目标图象。由于层次纹理图象需要较大的存储空间,并且在装入纹理缓冲时要花费大量时间,本文还提出一种基于压缩层次纹理图象的目标图象生成算法。层次纹理映射方法充分发挥图形硬件的功能,可以模拟粗糙物体表面的三维凹凸细节,这是传统纹理映射方法难以表现的。实验表明新算法是有效的,尤其适于处理那些与整个物体大小相比深度层次不太多的三维景物,如有表面凹凸纹理(浮雕、门窗等)的建筑物表面等,可以达到实时效果。
英文摘要: Image-based rendering (IBR) techniques have been proposed as an efficient way to generate novel views of real and synthetic objects. With traditional geometry-based rendering techniques, the time required in rendering an image increases with the geometric complexity of the scene. The cost of IBR techniques just depends on the resolution of the images and it can offer an easier modeling process at a higher realistic level. Since its initiation in the early 1990s, IBR has been developed very fast and widely used in many areas such as virtual reality, computer animation. This thesis restricts itself on the topic of image-based rending techniques. Some efficient IBR algorithms have been proposed to render novel views of complex virtual environment interactively, or in real time. In Chapter 2, we present an efficient depth-estimating-based IBR algorithm for novel view generation from multiple reference images in a static scene by combining forward mapping technique and inverse mapping technique. The method successfully fills the holes in the derived image in both cases where object surfaces are magnified in the novel view or occluded surfaces become visible. This algorithm is more efficient by the heuristic depth estimating technique proposed in this paper which gets displacement values used during backward mapping process from a single primary reference image under the observation that, the shape of space surfaces often changes smoothly. In comparison with usual forward mapping methods, this algorithm generates derived images with less errors and tackles the problems associated with multiple reference images. By taking advantages of epipolar line features and depth discontinuities in reference images, an efficient inverse warping algorithm is proposed in Chapter 3 for generating images of novel views by combining multiple reference images taken from different viewpoints. Because continuous segments determined by pairs of edge pixels at corresponding epipolar lines are order-kept, only pairs of edge pixels in the reference image are necessary to compute to obtain generalized disparity of all points in the desired image. As a result, significant acceleration could be made in the rendering process. Two accelerating techniques are presented in this algorithm to accelerate the hole filling process. This algorithm extends the reference images from projection of single connected surface in previously developed inverse warper to images captured from complex scene. In Chapter 4, an efficient IBR method is presented by taking full advantage of raphics card. The method can simulate the 3D details on surface of object successfully. The new approach, called layered texture mapping, consists of two parts. At first, an original texture with orthogonal displacements per pixel is decomposed into a series of new textures with each having a given displacement per pixel, called layered textures, or LT. Then, these LTs are used to render the novel view by conventional texture mapping. To avoid gaps in the rendered image, some pixels are to be interpolated and extended in the LTs according to the depth difference between two neighbor pixels in the original texture. As these TLs require much storage and therefore much time is required to install LTs into the texture buffer, a method is proposed to compress the LTs, and the corresponding rendering method is given. Experimental results show that the new method is efficient, especially in rendering those objects with a smaller depth range compared with their size, such as relief surfaces of building.
语种: 中文
内容类型: 学位论文
URI标识: http://ir.iscas.ac.cn/handle/311060/7508
Appears in Collections:中科院软件所

Files in This Item:
File Name/ File Size Content Type Version Access License
LW004426.pdf(1602KB)----限制开放-- 联系获取全文

Recommended Citation:
郑新. 基于图象的快速绘制技术的研究[D]. 中国科学院软件研究所. 中国科学院软件研究所. 2001-01-01.
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[郑新]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[郑新]‘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 © 2007-2017  中国科学院软件研究所 - Feedback
Powered by CSpace