基于图像绘制技术的研究
文献类型:学位论文
作者 | 胡金辉 |
学位类别 | 工学硕士 |
答辩日期 | 2001-05-01 |
授予单位 | 中国科学院自动化研究所 |
授予地点 | 中国科学院自动化研究所 |
导师 | 胡占义 |
关键词 | 基于图像的绘制 多同心拼图 相邻插值 离散漫游圆 虚拟现实 Image based rendering Multiple concentric mosaics Image interpolation system Discrete circle roaming virtual reality |
其他题名 | Research on Image Based Rendering Techniques |
学位专业 | 模式识别与智能系统 |
中文摘要 | 基于图像的绘制(Image Based Rendering,IBR)是近年来提出的一种全新的产生 真实感图像的方法,也是当前计算机图形学和计算机视觉最热门的研究方向之一。IBR 的基本思想是如何从已知图像中生成新的图像。它所追求的目标可以概括为:真实 性、实时性、实用性。近年来IBR技术的发展表明,添加多少几何信息、如何添加几 何信息以及如何简化全光线函数已成为IBR这一领域中越来越重要的研究课题。 为了满足IBR系统中图像绘制的真实性,本文对当前最有前途的一种IBR技术, 同心拼图[1]的深度进行了研究,并且在同心拼图的基础上提出了多同心拼图的方法。 多同心拼图法具有如下优点:1.操作简单、方便。只需将摄像机绕一固定中心、半径 不同的两个圆旋转拍摄两圈,产生两组同心拼图。2.可以方便快捷地估计出物体的深 度。后一点是非常重要的,因为只有计算出景物的深度才能产生具有真实感的图像。 多同心拼图法只需要通过简单的比例运算,不需要进行三维重建,因而大大简化了计 算,并提高了鲁棒性和可靠性(因为三维重建一般不鲁棒)。多同心拼图的方法实际 上是一种图像技术和几何信息相结合的方法,它采用的几何信息是图像之间的对应关 系,因此它的关键问题也在于对应。幸运的是,在多同心拼图中,由于特殊的极线约 束关系,对应问题大大简化。本文详细地讨论了图像之间的对应问题。多同心拼图法 不仅可以解决同心拼图中的图像真实感问题,更重要的是它可以解决文献[24]所提出的 大范围漫游采用的链式同心拼图中的重影问题。 本文还对IBR系统的实用性进行了研究。首先,本文分析了同心拼图法实用的困 难,然后在同心拼图的基础上提出了一种比较实用的IBR系统:相邻插值的IBR系 统。相邻插值的IBR系统的优点在于对实验设备要求比较低,并且数据量较小。另外 本文提出了采用离散漫游圆进行大范围漫游系统的尝试。和文献[24]中提出的链式同心 拼图相比较,离散漫游圆法具有较小的数据量,而且不需要精确知道各漫游圆之间的 运动参数。 实验证明了多同心拼图法、相邻插值法和离散漫游圆法的有效性。 |
英文摘要 | Image based rendering (IBR) is a novel technique of rendering photorealistic images, it is also one of the most active areas of computer graphics and computer vision. The basic idea of IBR is how to render new images from old ones, and its ultimate objectives are: photorealistic images, real time and low cost. The development of image based rendering has shown that the key problems in the field can be summarized as: how much geometric information should be used. how to use geometric information and how to simplify the plenoptic function To deal with the photorealistic problem in concentric mosaics(CM)[1], we proposed a new image based rendering method: multiple concentric mosaics(MCM), which can effectively eliminate the vertical distortion problem embedded in [1 ]. MCM technique has the following advantages: First, capturing MCM is simple and convenient. MCM are created by constraining the camera to rotate about a same center but different radius. Second, MCM can obtain depth information through simple calculation and effectively solve the vertical distortion problem embedded in CM. Actually, MCM technique can be classified as an IBR technique using implicit geometric information, the geometric information used here is correspondence, which is also the key problem of MCM. Fortunately, this problem can be substantially simplified due to the peculiar epipolar geometry between two images captured in MCM. In addition, we also discuss some practical ways of correspondence in detail. Not only can MCM technique solve the vertical distortion in CM, but also it can eliminate the annoying double image problem in concatenated concentric mosaics[24] This paper has also studied the practical problem in IBR. Based on CM, we proposed a practical system: Image Interpolation System(IIS). Compared with classic CM, IIS does not need to accurately control the motion of capturing device, and it has much smaller data size. This paper has also presented a practical Large - Scale Scene Roaming system: Discrete Circle Roaming System(DCRS). Compared with concatenated concentric mosaics in[24], DCRS has much smaller data and lower requirement of the control of camera motion. Final experiments showed that MCM. IIS and DCRS are effective and practical. |
语种 | 中文 |
其他标识符 | 600 |
源URL | [http://ir.ia.ac.cn/handle/173211/7337] ![]() |
专题 | 毕业生_硕士学位论文 |
推荐引用方式 GB/T 7714 | 胡金辉. 基于图像绘制技术的研究[D]. 中国科学院自动化研究所. 中国科学院自动化研究所. 2001. |
入库方式: OAI收割
来源:自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。