Scientific Visualization: The Visual Extraction of Knowledge from Data
Published on 20. January 2006
X, 434 pages
978-3-540-30790-7 (ISBN)
Description
One of the greatest scientific challenges of the 21st century is how to master, organize and extract useful knowledge from the overwhelming flow of information made available by today's data acquisition systems and computing resources. Visualization is the premium means of taking up this challenge. This book is based on selected lectures given by leading experts in scientific visualization during a workshop held at Schloss Dagstuhl, Germany. Topics include user issues in visualization, large data visualization, unstructured mesh processing for visualization, volumetric visualization, flow visualization, medical visualization and visualization systems. The book contains more than 350 color illustrations.
More details
Other editions
Additional editions
Georges-Pierre Bonneau | Thomas Ertl | Gregory M. Nielson
Scientific Visualization: The Visual Extraction of Knowledge from Data
Book
11/2005
Springer
€213.99
Shipment within 7-9 days
Content
Adaptive Contouring with Quadratic Tetrahedra.- On the Convexification of Unstructured Grids from a Scientific Visualization Perspective.- Brain Mapping Using Topology Graphs Obtained by Surface Segmentation.- Computing and Displaying Intermolecular Negative Volume for Docking.- Optimized Bounding Polyhedra for GPU-Based Distance Transform.- Generating, Representing and Querying Level-Of-Detail Tetrahedral Meshes.- Split' N Fit: Adaptive Fitting of Scattered Point Cloud Data.- Ray Casting with Programmable Graphics Hardware.- Volume Exploration Made Easy Using Feature Maps.- Fantastic Voyage of the Virtual Colon.- Volume Denoising for Visualizing Refraction.- Emphasizing Isosurface Embeddings in Direct Volume Rendering.- Diagnostic Relevant Visualization of Vascular Structures.- Clifford Convolution and Pattern Matching on Irregular Grids.- Fast and Robust Extraction of Separation Line Features.- Fast Vortex Axis Calculation Using Vortex Features and Identification Algorithms.- Topological Features in Vector Fields.- Generalizing Focus+Context Visualization.- Rule-based Morphing Techniques for Interactive Clothing Catalogs.- A Practical System for Constrained Interactive Walkthroughs of Arbitrarily Complex Scenes.- Component Based Visualisation of DIET Applications.- Facilitating the Visual Analysis of Large-Scale Unsteady Computational Fluid Dynamics Simulations.- Evolving Dataflow Visualization Environments to Grid Computing.- Earthquake Visualization Using Large-scale Ground Motion and Structural Response Simulations.
Generating, Representing and Querying Level-Of-Detail Tetrahedral Meshes (p. 79-80)
Leila De Floriani1,2 and Emanuele Danovaro1
1 Department of Computer Science, University of Genova, Genova, Italy
2 Department of Computer Science, University of Maryland, College Park, MD, USA
Summary. In this paper, we survey techniques for building, encoding and querying Level-
Of-Detail (LOD) models of three-dimensional scalar fields based on a domain decomposition
into tetrahedral meshes. We focus on continuous LOD models, and we classify them into
unstructured (irregular) and regular nested LOD models depending on the mesh subdivision
pattern and on the distribution of the data points.Within each class, we review data structures,
construction algorithms, as well as techniques for extracting adaptively refined field representations
from an LOD model.
1 Introduction
Level-Of-Detail (LOD) models have been proposed to control and adapt the accuracy in the representation of large-size volume data sets. LOD models encode in a compact data structure the steps performed by a refinement process applied to a coarse representation of a scalar field, or by a decimation process applied to a full- resolution representation. A large numbers of simplified meshes can be extracted from an LOD model, in which the resolution (i.e., the density of the cells) of the simplified mesh may vary in different parts of the field domain, or in the proximity of interesting field values. The extraction of a simplified representation from an LOD model is called a selective refinement. The challenge in designing an LOD model is represented by the trade-off between the efficiency of the selective refinement algorithms and the storage cost of the representation.
This paper reviews techniques proposed in the literature for encoding, generating and performing selective refinement on an LOD model. We focus on so-called continuous LOD models, from which a virtually continuous simplified adaptive representations can be extracted. Discrete (non-continuous) LOD models consist of a (usually small) collection of representations at different LODs and only representations of the scalar field at uniform resolutions can be extracted from them [6]. The remainder of this paper is organized as follows. Section 2 introduces some background notions on tetrahedral meshes and discusses data structures for encoding them. Section 3 introduces the basic elements of an LOD model and the most common update operations through which an LOD model is generated. Section 4 reviews incremental refinement and coarsening techniques used to generate unstructured LOD models. Section 5 discusses data structures for encoding unstructured LOD models. Section 6 reviews techniques for encoding and querying nested LOD models. Section 7 presents some comparisons of performances of unstructured and nested LOD models in extracting adaptively-refined meshes, and discusses some open research issues.
2 Background
A volume data set S consists of a set V of points in the three-dimensional Euclidean space, and of one or several field values associated with the points of V. The points in V can be regularly spaced, i.e., they are the vertices of a regular, rectilinear grid, or irregularly spaced. In the former case, we will call S a structured, or a regular, data set, while, in the latter case, we will call it an unstructured, or an irregular data set. A tetrahedral mesh S is a connected set of tetrahedra such that the union of all tetrahedra in S covers a domain D in 3D space and any two distinct tetrahedra have disjoint interiors. A tetrahedral mesh Sis called a conforming mesh if the intersection of the boundaries of any two tetrahedra s1 and s2 of S, which have a non- empty intersection, consists of lower dimensional simplexes (vertices, edges, or triangles) that belong to the boundary of both s1 and s2. Conforming meshes have a wellde .ned combinatorial structure in which each tetrahedron is adjacent to exactly one other tetrahedron along each of its faces. This is important when a tetrahedral mesh is used as a decomposition of the domain of a volume data set.
We call nested meshes those meshes which are defined by the uniform subdivision of a tetrahedron into scaled copies of it. In particular, we will consider nested regular meshes, in which the vertices are a subset of the vertices of a regular grid. A mesh which is not nested is called irregular, or unstructured. A mesh is called stable if the tetrahedra forming it satisfy some measure of non-degeneracy. Measures commonly used in the .nite element literature are the circumradius- to-shortest-edge ratio (where the circumradius is the radius of the circumsphere of a tetrahedron), and the minimum solid angle associated with a tetrahedron [39].
Leila De Floriani1,2 and Emanuele Danovaro1
1 Department of Computer Science, University of Genova, Genova, Italy
2 Department of Computer Science, University of Maryland, College Park, MD, USA
Summary. In this paper, we survey techniques for building, encoding and querying Level-
Of-Detail (LOD) models of three-dimensional scalar fields based on a domain decomposition
into tetrahedral meshes. We focus on continuous LOD models, and we classify them into
unstructured (irregular) and regular nested LOD models depending on the mesh subdivision
pattern and on the distribution of the data points.Within each class, we review data structures,
construction algorithms, as well as techniques for extracting adaptively refined field representations
from an LOD model.
1 Introduction
Level-Of-Detail (LOD) models have been proposed to control and adapt the accuracy in the representation of large-size volume data sets. LOD models encode in a compact data structure the steps performed by a refinement process applied to a coarse representation of a scalar field, or by a decimation process applied to a full- resolution representation. A large numbers of simplified meshes can be extracted from an LOD model, in which the resolution (i.e., the density of the cells) of the simplified mesh may vary in different parts of the field domain, or in the proximity of interesting field values. The extraction of a simplified representation from an LOD model is called a selective refinement. The challenge in designing an LOD model is represented by the trade-off between the efficiency of the selective refinement algorithms and the storage cost of the representation.
This paper reviews techniques proposed in the literature for encoding, generating and performing selective refinement on an LOD model. We focus on so-called continuous LOD models, from which a virtually continuous simplified adaptive representations can be extracted. Discrete (non-continuous) LOD models consist of a (usually small) collection of representations at different LODs and only representations of the scalar field at uniform resolutions can be extracted from them [6]. The remainder of this paper is organized as follows. Section 2 introduces some background notions on tetrahedral meshes and discusses data structures for encoding them. Section 3 introduces the basic elements of an LOD model and the most common update operations through which an LOD model is generated. Section 4 reviews incremental refinement and coarsening techniques used to generate unstructured LOD models. Section 5 discusses data structures for encoding unstructured LOD models. Section 6 reviews techniques for encoding and querying nested LOD models. Section 7 presents some comparisons of performances of unstructured and nested LOD models in extracting adaptively-refined meshes, and discusses some open research issues.
2 Background
A volume data set S consists of a set V of points in the three-dimensional Euclidean space, and of one or several field values associated with the points of V. The points in V can be regularly spaced, i.e., they are the vertices of a regular, rectilinear grid, or irregularly spaced. In the former case, we will call S a structured, or a regular, data set, while, in the latter case, we will call it an unstructured, or an irregular data set. A tetrahedral mesh S is a connected set of tetrahedra such that the union of all tetrahedra in S covers a domain D in 3D space and any two distinct tetrahedra have disjoint interiors. A tetrahedral mesh Sis called a conforming mesh if the intersection of the boundaries of any two tetrahedra s1 and s2 of S, which have a non- empty intersection, consists of lower dimensional simplexes (vertices, edges, or triangles) that belong to the boundary of both s1 and s2. Conforming meshes have a wellde .ned combinatorial structure in which each tetrahedron is adjacent to exactly one other tetrahedron along each of its faces. This is important when a tetrahedral mesh is used as a decomposition of the domain of a volume data set.
We call nested meshes those meshes which are defined by the uniform subdivision of a tetrahedron into scaled copies of it. In particular, we will consider nested regular meshes, in which the vertices are a subset of the vertices of a regular grid. A mesh which is not nested is called irregular, or unstructured. A mesh is called stable if the tetrahedra forming it satisfy some measure of non-degeneracy. Measures commonly used in the .nite element literature are the circumradius- to-shortest-edge ratio (where the circumradius is the radius of the circumsphere of a tetrahedron), and the minimum solid angle associated with a tetrahedron [39].
