Wavelet-based multiresolution with n√2 subdivision

被引：7

作者：

Linsen, L ^{[1
]}

Pascucci, V

Duchaineau, MA

Hamann, B

Joy, KI

机构：

[1] Univ Calif Davis, Dept Comp Sci, CIPIC, Davis, CA 95616 USA

[2] Lawrence Livermore Natl Lab, Data Sci Grp, CASC, Livermore, CA 94550 USA

来源：

COMPUTING | 2004年 / 72卷 / 1-2期

关键词：

multiresolution modeling; (time-varying) volume modeling; subdivision; B-spline wavelets; data approximation;

D O I：

10.1007/s00607-003-0052-0

中图分类号：

TP301 [理论、方法];

学科分类号：

081202 ;

摘要：

Multiresolution methods are a common technique used for dealing with large-scale data and representing it at multiple levels of detail. We present a multiresolution hierarchy construction based on (n)root2 subdivision, which has all the advantages of a regular data organization scheme while reducing the drawback of coarse granularity. The (n)root2-subdivision scheme only doubles the number of vertices in each subdivision step regardless of dimension n. We describe the construction of 2D, 3D, and 4D hierarchies representing surfaces, volume data, and time-varying volume data, respectively. The 4D approach supports spatial and temporal scalability. For high-quality data approximation on each level of detail, we use downsampling filters based on n-variate B-spline wavelets. We present a B-spline wavelet lifting scheme for (n)root2-subdivision steps to obtain small or narrow filters. Narrow filters support adaptive refinement and out-of-core data exploration techniques.

引用

页码：129 / 142

页数：14