A Hybrid Representation of 3D Objects for Fast Rendering and Compression
Date Issued
2006
Date
2006
Author(s)
Lo, Wan-Yen
DOI
en-US
Abstract
We introduce a new representation for 3D objects, aiming at real-time rendering and high compression ratio, so that the virtual objects could be more applicable to many real-world applications. Since both pure model-based and pure image-based approaches have particular merits, we try to take advantages from them both and to make the 3D geometry and acquired images complement with each other. To simplify the geometry for efficient rendering, we propose the idea of billboard clusters, which is a combination of billboard clouds and layered impostors. The simplified 3D model is then texture-mapped with the acquired images to achieve photo-realistic rendering effects. Furthermore, before texture mapping, the view-dependent lighting effects, shading and specular components, should be removed from the image sequence to ensure color-consistent outcome. We then formulate the diffuse-specular separation problem in a Bayesian framework, and solve the optimum albedo of each pixel with the maximum-a-posteriori technique. As a consequence, the the object is even able to be relighted under arbitrary illumination conditions.
Subjects
影像式繪圖
影像式表示法
三維模型化簡
分離漫反射與鏡面反射成分
image-based rendering
image-based representation
model simplification
diffuse-specular separation
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-95-R93922050-1.pdf
Size
23.31 KB
Format
Adobe PDF
Checksum
(MD5):83ea3d71bb7c64575bb737f28214e3ea