CASIA OpenIR  > 毕业生  > 硕士学位论文
Alternative TitleResearch on Image Based Modeling and Rendering
Thesis Advisor胡占义
Degree Grantor中国科学院研究生院
Place of Conferral中国科学院自动化研究所
Degree Discipline模式识别与智能系统
Keyword基于图像的建模与绘制 人脸 图像变形 光照 三维重建 Ibmr Faces Image Morphing Illumination 3d Reconstruction
Abstract基于图像的建模与绘制技术(Image Based Modeling and Rendering,IBMR)是近年来兴起的一种综合性的多学科交叉技术,它应用了计算机视觉、图像处理和计算机图形学等最新的研究成果。它所追求的目标可以概括为:真实性、实时性和实用性。本文的研 究主要围绕人脸的绘制和建模进行,主要工作有: 1.实现了一个自动提取人脸特征点的原型系统。在应用Gabor滤波器对原始图像作 多尺度,多方向的滤波来进行特征提取的基础上,对特征点匹配的几种相似函数 一一作了比较。实验表明Gabor滤波器是一种很好的特征提取工具。在原有的理 论基础上,我们提出了用特征点间的不变量来进一步提高特征点匹配鲁棒性的方 法。自动提取特征点并建立特征点之间对应关系之后,我们应用多层次自由形变 的方法实现了图像变形。对于非均匀变换采用三角化的方法来求得非特征点的变 换系数,并利用对原图像作前变换的手段来得到视觉上比较完美的图像。 2.光照是计算机视觉中的许多问题,例如,图像绘制、人脸识别、物体的重建等, 都必须考虑的一个重要因素。研究在某种简化的光照模型下图像的重绘技术,以 及在此基础上研究物体的三维重建问题一直是计算机视觉的一个很重要的研究内 容。我们提出了一种新的用于人脸在不同光照下重绘的方法,实验表明,该方法 可以取得比较满意的重绘效果。应用变化光照下的多幅图像也可以对于物体进行 三维重建。我们把物体的重建分成两个过程。从二维到2.5维的重建,2.5维是二 维图像和三维物体之间的中间表示层次,在Lambert光照模型下利用一幅或者多 幅图像求得物体形状的一个重要参数--法向量。计算出物体的法向量后,对于 如何从物体的法向量重建物体的三维形状,我们提出了一种线性迭代的算法,实 验结果表明这种算法具有很好的收敛性和鲁棒性。
Other AbstractImage Based Modeling and Rendering (IBMR) is a new emerging technology, it has found many promising applications in different fields, in particular, in Virtual Reality. Its three goals are: photorealistic images, real time rendering and low cost. The thesis is focused on the rendering and modeling of human faces, and can be summarized as follows: 1. A prototype system is implemented. It can automatically locate specific facial points on a face image. Gabor filter with different frequencies and orientations are used for feature extraction, then three similarity functions are compared. Experimental results show that Gabor filter is indeed a powerful tool in feather extraction. In addition, an invariant-based technique is used to select the final feature correspondence, and is showed to be effective. After the establishment of facial control point correspondences, Multilevel of Free Form Deformation(MFFD) is used in image morphing, where the control function at every point is computed using the method of triangulation in non-uniform metamorphosis and a pre-shifting step is engaged to reduce the distortion in the morphing. 2. Due to the adverse effect of illumination on image rendering, face recognition and reconstruction, researchers in computer vision have paid special attention to the study on image rendering with varying illumination and shape from shading. A new approach for photorealistic rendering of faces under arbitrary illumination is presented in this work, and experimental results validate it. In addition, we also investigate a technique for object reconstruction using several images under varying illumination. The technique is divided into two steps. In the first step, a 2.5D shape, i.e., normal of the shape at each point, is obtained. Then in the second step, a new linear-iterative method is proposed to upgrade the shape to the 3D one. The method is shown to be of good convergence and robustness.
Other Identifier684
Document Type学位论文
Recommended Citation
GB/T 7714
王城峰. 基于图像建模与绘制技术研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2003.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[王城峰]'s Articles
Baidu academic
Similar articles in Baidu academic
[王城峰]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[王城峰]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

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