Chandrajit Bajaj

DATA VISUALIZATION TECHNIQUES

Edited by Chandrajit Bajaj

DATA VISUALIZATION TECHNIQUES, Edited by C. Bajaj

c 1998 John Wiley & Sons Ltd
Preface
Data visualization concerns the manipulation of sampled and computed data for com-
prehensive display. The goal of the visualization to bring to the user a deeper un-
derstanding of the data as well as the underlying physical laws and properties. Such
visualization may be used to enlighten a physicist on the complex interaction between
electrons, to guide the medical practitioner in a surgery situation, or simply to view
the surface of a planet which has never been seen by human eyes.
Through the presentation of massive amounts of data as images, we allow the vi-
sualization user to rapidly prune useless information, focus on necessary information,
and comprehend the science behind the data. Interaction with data brings another
level of understanding. Static images can be misleading and mask important features
of the data. Motion in visualization brings out hidden features which are inherently
dynamic. Interactive manipulation and control of visualization is an important tool
which allows scientists to more quickly focus on the region of interest. In environments
which are immersive, the motion is critical, to the point that delays or inconsistencies
can make the viewer ill. In this case there is a desire to bound response time using
time-critical techniques.
The important aspects of interactive visualization can be broken down into three
categories:
Computation - the ability to speedily compute a visualization. This may include
computing a polygonal approximation to an isosurface of a scalar function, or the
computation of a particle trace through a time-dependent vector eld, or any action
which requires extracting an abstract object or representation from the data being
examined.
Display - the ability to quickly display the computed visualization. Display encom-
passes both computed visualizations as listed above, as well as direct display meth-
ods such as volume visualization and ray tracing.
Querying - the ability to interactively probe a displayed visualization for the purpose
of further understanding on a ne scale what is begin displayed on a coarser scale.
In this book we propose to address the important aspects of interactivity in the
visualization of scalar, vector and tensor eld data. In addition, the book shall address
data structures and algorithmic techniques for ecient computation and visualization
in the time domain.

DATA VISUALIZATION TECHNIQUES, Edited by C. Bajaj

c 1998 John Wiley & Sons Ltd
2 Chandrajit Bajaj

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

1

Visualization Paradigms
Chandrajit L. Bajaj
University of Texas, Austin

ABSTRACT

A wide variety of techniques have been developed for the visualization of scalar,
vector and tensor eld data. They range from volume visualization, to isocontour-
ing, from vector eld streamlines or scalar, vector and tensor topology, to function
on surfaces . This multiplicity of approaches responds to the requirements emerg-
ing from an even wider range of application areas such as computational uid
dynamics, chemical transport, fracture mechanics, new material development,
electromagnetic scattering/absorption, neuro-surgery, orthopedics, drug design.
In this chapter I present a brief overview of the visualization paradigms currently
used in several of the above application areas. A major objective is to provide
a roadmap that encompasses the majority of the currently available methods
to allow each potential user/developer to select the techniques suitable for his
purpose.

1.1 Introduction
Typically, informative visualizations are based on the combined use of multiple tech-
niques. For example gure 1.1 shows the combined use of isocontouring, volume ren-
dering and slicing to highlight and compare the internal 3D structure of three dierent
vorticity elds. For a detailed description of each of the approaches we make reference
to subsequent chapters in this book and previously published technical papers and
books [Bow95, Cle93, HU94, KK93, NHM97, REE+ 94, Wat92].
DATA VISUALIZATION TECHNIQUES, Edited by C. Bajaj
c 1998 John Wiley & Sons Ltd
4 Bajaj

Figure 1.1 The combined display of isocontours, slicing and volume rendering used to
highlight the 3D structure of vorticity elds.

Figure 1.2 Two volume renderings showing snapshots of wind speed in a global climate
model.
1.2 Volume Rendering
Volume rendering is a projection technique that produces image displays of three-
dimensional volumetric data (see g. 1.2). Its main characteristic is the production of
view-dependent snapshots of volumetric data, rather than the extraction of geometric
information such as isocontouring.
Chapter 2 surveys alternate volume rendering algorithms reported in the literature.
Two main classes of approaches that have been developed dier mainly on the order

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

Visualization Paradigms 5
of projection of the volume cells. Secondary distinctions arise from the dierences in
color accumulation and composition techniques to produce the nal image.
Forward projection techniques traverse the volume (object space approach) pro-
jecting and display each volumetric cell or voxel. This approach takes advantage
of graphics hardware acceleration by selecting appropriate drawing primitives to
approximate the voxel image.
Backward projection techniques traverse the image (image space approach) and
cast through the data volume, one light ray per pixel, accumulating color intensities
along the ray to determine the nal pixel color.
Cell projection and splatting are both forward projection techniques. In cell pro-
jection, the cells of the data volume are traversed and their images computed by
subdivision into a polygonal approximation. In splatting, the samples of the volume
are traversed and their contribution to the nal image is computed by convolution with
a reconstruction kernel. Cell projection technique can be optimized by taking advan-
tage of the spatial coherence of the volume cells both in the case of regular grids and
in the case of unstructured meshes. Splatting has been shown to be a fast technique
for hardware assisted scalar volume visualization, and was extended to vector elds
(see details in Chapter 5) Additional splatting techniques are developed for texture
based visualization of velocity elds in the vicinity of contour surfaces (see details in
Chapter 6)
Backward projection methods are accelerated by exploiting the coherence between
adjacent rays. This idea has been implemented in a number of approaches using:
(i) adaptive sampling along the rays depending on the \importance" of dierent re-
gions(ii) templating the paths of parallel rays through regular grids,(iii) bounding with
simple polyhedra signicant regions that give the main contribution to the output im-
age, or (iv) maintaining the front of propagating rays through irregular grids. The
high computational cost of volume rendering in the spatial domain can sometimes be
replaced by an asymptotically faster computation in the frequency domain [Lev92,
Mal93, TL93].

1.3 Isocontouring
Isocontouring is the extraction of constant valued curves and surfaces from 2d and 3d
scalar elds. Interactive display and quantitative interrogation of isocontours helps in
determining the overall structure of a scalar eld (see g. 1.3) and its evolution over
time (see g. 1.4).
Chapter 3 surveys the most commonly used isocontouring algorithms along with
recent improvements that permit rapid evaluation of multiple isocontour queries, in
an interactive environment. Traditional isocontouring techniques examine each cell of
a mesh to test for intersection with the isocontour of interest. Accelerated isocontour-
ing can be achieved by preprocessing the input scalar eld both in its domain (the
geometry of the input mesh) and in its range (the values of the sampled scalar eld).
One the one hand, one takes advantage of the known adjacency information of mesh
cells (domain space optimization). Given a single cell c on an isocontour component
one can trace the entire isocontour component from c, by propagating from cell to cell
16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book
6 Bajaj

using inter-cell adjacency. This reduces the search for isocontour components from a
search in the entire input mesh to a search in a much smaller subset called the seed set.
A seed set is a subset of the mesh cells which has at least one cell on each connected
component of each isocontour. From this typically very small seed set of mesh cells
one searches for starting cells for each component of the desired isocontour and then
applies contour propagation through cell adjacencies.
On the other hand, one independently optimizes the search for isocontours exploit-
ing the simplicity of the range of the scalar eld (range space optimization). The
values of the eld are scalars that in range space form an interval. Within each cell
of the mesh (or of mesh cells of the seed set) the scalar eld usually has a small
continuous variation that can be represented in range space as a (small) subinterval.
The isocontour computation is hence reduced in range space to the search for all the
segments that intersect the currently selected isovalue w. This search can be optimally
performed using well known interval tree or segment tree data structures.

Figure 1.3 Skin and bone head models extracted as two dierent isocontours from the
same volumetric MRI data of the Visible Human female.

Figure 1.4 Three isocontours of wind speed that show the time evolution of air dynamics
in a global climate model.

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

Visualization Paradigms 7
1.4 Flow Visualization
Visualization of vector elds is generally more complicated than visualizing scalar
elds due to the increased amount of information inherent in vector data. Clearly
vector data can be contracted to scalar quantities, for example by computation of
vector magnitude, scalar product with a given vector, or magnitude of vorticity. In
this case, scalar visualization techniques such as isosurfaces and volume rendering can
be applied. Additional approaches to visualization of vector elds include iconogra-
phy, particle tracking, and qualitative global ow visualization techniques. Chapter 5
reviews ow visualization techniques while Chapter 6 describes more in detail the ap-
proaches designed to take advantage of currently existing graphics hardware to increase
performance. For additional detail, refer to the papers cited in these two chapters.
Particle tracking or advection techniques are based on following the trajectory of a
theoretically massless particle in a ow. In its simplest form, the path traversed by a
particle in a steady ow is called a streamline. If the ow is unsteady, or time-varying,
the path followed by a particle over time is called a path line. A curve resulting from
a number of particles emitted at regular or irregular intervals from a single source
is called a streak line. Numerical techniques commonly used for evaluating the above
equation include Euler and Runge-Kutta methods. In the case of incompressible ow, a
single stream-function in 2D can be constructed such that the contours of the stream-
function are streamlines of the vector eld. In 3D, a pair of dual stream functions
is required, and streamlines will occur as the intersections of isocontours of the two
functions [KM92].
Particle tracking techniques may also be extended by grouping multiple particles
together to form a stream ribbon, stream surface, stream tube or ow volume. Global
techniques such as Line Integral Convolution present a qualitative view of the vector
eld which presents intuitively meaningful visualizations for the user. Flow \probes"
may be placed at user-specied or computed locations to reveal local properties of the
ow eld such as direction, speed, divergence, vorticity, etc. Properties are mapped to a
geometric representation called an icon. The complexity of the icon increases with the
amount of information that it is designed to represent. Representing curl (vorticity),
which is itself a vector eld with additional physical meaning, can be achieved by a
cylindrical icon with candystriping to indicate both the direction and magnitude of
vorticity.

1.5 Quantication

In the quest for interrogative visualization [Baj88], in which the user can not only see
the data, but navigate and query for increased understanding, the ability to quantify
and perform volumetric measurements is vital. Another challenge to visualization is
to give quantitative information concerning time dependent studies and time-varying
structures (e.g. ow). In the study of paralysis, researchers are constructing models of
spinal cords and regions of damaged cord from histological samples. Figure 1.5 (left)
is an example of a histological slice of an injured rat spine. In Figure 1.5 (right),
the damaged region has been reconstructed as a surface, and is visualized along with
orthogonal slices of the 3D histological specimen. Traditionally,spinal damage has been
16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book
8 Bajaj

modelled as an expanding cylindrical region. The ability to more accurately dene the

region of damage and measure the surface area and volume of the region are promising
tools in developing a greater understanding of cysts and how they develop.
Chapter 7 reviews tissue classication techniques using local reconstructions of band
limited samplings, and Bayesian statistics. Such classication provides the means to
accurately identify (for isocontouring and volume rendering), and quantify the relevant
substructures of three dimensional images.

Figure 1.5 (Left) Histological sample of a rat spine. (Right) Reconstruced spinal lesion
within slices of 3D histological volume.
Chapter 4 reviews the shape analysis and visualization of free form surface mod-
els used in computer aided geometric design and computer graphics. The analysis
tools prove essential to detecting surface imperfections aswell as higher order inter-
patch smoothness. Related research on free-form surfaces visualization are addressed
in [BR94, BBB+ 97].
Three special cases of volumetric quantication which are prevalent in data visual-
ization applications apply to the following data types:
 Contours - surfaces which are created through isocontouring of scalar data
 Slices - surfaces which are formed by tiling multiple planar cross sections of objects
 Union of Balls - also known as the solvent accessible surface and common in
molecular visualization
Contour Quantication
Bajaj, Pascucci and Schikore [BPS97] introduce the systematic quantication of met-
ric properties of volumetric data and the relative isocontours. Given an isovalue w
one can compute the surface area of the corresponding isosurface, the volume of the
inside region or any other metric property (also called signature) function of w. The
plot of the signatures gives rise to an interface that drives the user in the direct
selection of interesting isovalues. Figure 1.6 shows the direct selection of noiseless iso-
surfaces corresponding to skin and bone tissues which correspond to the maxima of
the gradient-weighted area signature.
Sliced Data
Objects are frequently reconstructed from serial sections [BCL96b, BCL96a]. In this
case, volume properties can be accurately computed using the following equation for
16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book
Visualization Paradigms 9

Figure 1.6 Three isosurfaces of the same volumetric MRI scan. The vertical bars in the
spectrum interface (top) mark the selected isovalues.
prismatoids, a triangular tiling of two parallel contours: V = h6 (B1 + 4M + B2 ) where
B1 is the area of lower base, B2 is the area of upper base, M is the area of the
midsection joining the bases, and h is the separation between the contours. With n
P equallyPspaced, the composite volume computation results
parallel slices of contours
in: V 0 = h6 (B1 + 4 1n?1 Mi + 2 2n?1 Bi + Bn ).
Union of Balls
The geometric, combinatorial and quantitative structure of the union of a set of balls
has been presented by [Ede95, DE95]. The union of balls model is equivalent to the
space lling model used to represent molecules where each atom is approximated by
a ball with a relative van der Waals radius. Deeper insight on the properites of a
molecule in solution is provided by the computation of the Solvent Accessible Surface
and the Solvent Excluded Surface [SSO96]. The two surfaces are dened by idealizing
the solvent molecule (e.g. water) as a single ball and computing the boundary of the
region that can be accessed by the solvent center (Solvent Accessible oset of the Union
of Balls model of the molecule) or the boundary of the region that cannot be reached
by any point of the solvent (Solvent Excluded). On the basis of the union of balls model

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

10 Bajaj

exact representation of both the surfaces can be computed eciently [BLMP97] (see
g. 1.7).

Figure 1.7 (top) Union of Balls, Solvent Accessible and Solvent excluded surfaces of the
Nutrasweet molecule with respect to the same solvent. (bottom) Solvent excluded surface of
the Gramicidin molecule with respect to three increasing solvent radii.

1.6 Data Reduction

Mesh reduction or simplication refers to a broad category of techniques designed to
trade space and complexity for accuracy in representation of a surface or volume mesh.
Like isocontouring mesh reduction is an algorithmic approach used to preprocess the
input dataset to make more suitable for display or analysis queries. The dierence
is that while isocontouring extracts an interesting feature like a particular isosurface
from a volumetric dataset, mesh reduction is meant to generate a reduced version
of the volume itself to speedup postprocessing. Figure 1.8 demonstrates mesh reduc-
tion applied to 2D functional data, in this case a slice of MRI data. Related results
come from several research communities, including Geographical Information Systems
(GIS), Computational Fluid Dynamics (CFD), and Virtual Environments/Virtual Re-
ality (VE/VR). Each community has much the same goal for achieving interactivity
with very large sets of data. An initial classication of techniques can be made by
distinguishing between static simplication, in which a single resolution output is com-
puted from a high-resolution input based on given simplication criteria, and dynamic
16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book
Visualization Paradigms 11

Figure 1.8 Original data (left) Reduced Triangulation (center) Reduced Image (right).
simplication, in which hierarchical triangulations are computed, from which reduced
triangulations can be rapidly extracted depending on time dependent metrics such
as distance from the viewer, location in the eld of view, and approximation error
tolerance.
Height-eld reduction: A driving application for reduction of height-elds is GIS. A
wide range of techniques are based on extraction of key points or edges from the orig-
inally dense set of points, followed by a constrained Delaunay triangulation [DFP85,
FFNP84, FL79, PDDT94, Tsa93, WJ92]. Silva, et. al [SMK95] uses a greedy method
for inserting points into an initially sparse mesh, reporting both better and faster re-
duction compared to a freely available terrain reduction tool. A survey by Lee [Lee91]
reviews methods for computing reduced meshes by both point insertion and point dele-
tion. Bajaj and Schikore [BS95, BS96a] developed practical techniques for measuring
the local errors introduced by simplication operations and bound the global error
accumulated by multiple applications. Their techniques begin with simple scalar elds
and extend easily to multi-valued elds and dened on arbitrary surfaces. Geometric
error in the surface as well as functional error in the data are bounded in a uniform
manner. Topology preserving, error-bounded mesh simplication have also been ex-
plored [BS97]. Figure 1.9 demonstrates geometric mesh reduction while Figure 1.8
demonstrates mesh reduction applied to 2D functional data.
Geometry reduction: Geometric mesh reduction has been approached from several
directions. In the reduction of polygonal models, Turk [Tur92] used point repulsion
on the surface of a polygonal model to generate a set of vertices for retriangulation.
Schroeder, et al. [SZL92] decimate dense polygonal meshes, generated by Marching
Cubes [LC87], by deletion of vertices based on an error criteria, followed by local retri-
angulation with a goal of maintaining good aspect ratio in the resulting triangulation.
Errors incurred from local retriangulation are not propagated to the simplied mesh,
hence there is no global error control. Rossignac, et al.[RB93] uses clustering and
merging of features of an object which are geometrically close, but may not be topo-
logically connected. In this scheme, long thin objects may collapse to an edge and
small objects may contract to a point. Hamann [Ham94] applies a similar technique
in which triangles are considered for deletion based on curvature estimates at the
vertices. Reduction may be driven by mesh resolution or, in the case of functional

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

12 Bajaj

Figure 1.9 Three snapshots of geometric mesh simplication of an engine block
surfaces, root-mean-square error. He, et al. [HHK+ 95] perform mesh reduction by
volume sampling and low-pass ltering an object. A multi-resolution triangle mesh is
extracted from the resulting multi-resolution volume buer using traditional isosur-
facing techniques. Hoppe, et al. [HDD+ 93] perform time-intensive mesh optimization
based on the denition of an energy function which balances the need for accurate
geometry with the desire for compactness in representation. The level of mesh sim-
plication is controlled by parameters in the energy function which penalizes meshes
with large numbers of vertices, as well as a spring constant which helps guide the
energy minimization to a desirable result.
In [Hop96], Hoppe introduces Progressive Meshes, created by applying optimization
with the set of basic operations reduced to only an edge contraction. Scalar attributes
are handled by incorporating them into the energy function. Ronfard, et. al [RR96]
also apply successive edge contraction operations to compute a wide range of levels-of-
detail for triangulated polyhedra. Edges are extracted from a priority queue based on
a computed edge cost such that edges of lesser signicance are removed rst. Cohen,
et al. [CVM+ 96] introduce Simplication Envelopes to guide mesh simplication with
global error bounds. Envelopes are an extension of oset surfaces which serve as an ex-
treme boundary for the desired simplied surface. Lindstrom, et. al[LKR+ 96] impose
a recursive triangulation on a regular terrain and compute preprocessing metrics at
various levels of resolution which permits real-time adaptive triangulation for interac-
tive y-through. Funkhouser et al. describe adaptive display algorithms for rendering
complex environments at a sustained frame rate using multiple levels of detail [FS93].
Delaunay techniques for static simplication have been extended to create hierar-
chies of Delaunay triangulations from which a simplied mesh can be extracted on the
y [dBD95]. Successive levels of the hierarchy are created by deleting points from the
current level and retriangulating according to the Delaunay criteria, giving the hier-
archical structure of a directed acyclic graph (DAG). Puppo improves on the approach
of de Berg by augmenting the DAG with information on which triangles between suc-
cessive resolutions are overlapping [Pup96]. With this information, the problem of
extracting a triangulation from the DAG is simplied and requires no backtracking.
It is shown that for a given triangulation criteria, the optimal triangulation satisfying
the criteria which is embedded in the DAG can be extracted in optimal time. Cohen

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

Visualization Paradigms 13

(a) (b)

Figure 1.10 (a) Arrow plot of a two dimensional vector eld (b) Streamline along focus in
the vortex core surrounded by nearby streamlines.
and Levanoni adopt a tree representation for hierarchical Delaunay triangulation and
demonstrate techniques for maintaining temporal coherence between successive trian-
gulations [COL96]. The technique is demonstrated for relatively sparse terrains, and
it remains to be seen whether the constraints imposed by a tree representation will
restrict simplication for dense triangulated terrains.
Wavelets [Mal89, Dau92] have been utilized for their multiresolution applications
in many areas of computer graphics and visualization [SDS96], including image com-
pression [DJL92a], surface description [DJL92b, EDD+ 95, CPD+ 96], tiling of con-
tours [Mey94] and curve and surface editing [FS94, ZSS97]. A number of multiresolu-
tion volume hierarchies have been proposed for developing adaptive volume rendering
and isocontouring [Mur92, Mur95, CDM+ 94, WV94].

1.7 Topology
Field topology refers to the analysis and classication of critical points and computa-
tion of relationships between the critical points of eld data [Del94]. Computation and
display of eld topology can provide a compact global view of what is otherwise a very
large set of data. Techniques such as volume rendering and line integral convolution
provide qualitative global views of eld topology.
Vector Topology: Given a continuous vector eld, the locations at which the vector
becomes zero are called critical points. Analysis of the critical points can determine
behavior of the vector eld in the local region. In a 2D vector eld, critical points are
classied into sources, sinks, and saddles (see Figure 1.11(a)), with both spiral and
non-spiral cases for sources and sinks [HH90, HH89, GLL91]. In 3D, additional critical
16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book
 14 Bajaj

points include the spiral-saddle [HH90], which is useful for locating vortex cores, as
shown in Figure 1.10.

Spiral Source Saddle Source Maxima Minima Regular Saddle

Degenerate Saddle Constant

Spiral Sink Center Sink

Figure 1.11 (a) Classication of Vector Field zeroes. (b) Critical point classication for
Scalar Field.

Figure 1.12 Two examples of scalar topology of 2D (left) and 3D (right) scalar elds.
The 2D case shows the scalar topology displayed over a color-map of the density in a pion
collision simulation. The 3D case is that of the scalar topology diagram of the wave function
computed for a high potential iron protein.
Scalar Topology: Scalar eld topology can be viewed as a special case of vector eld
topology, where the vector eld is given by the gradient of the scalar function [BS96b].
Critical points in a scalar eld are dened by a zero gradient, and can be classied into
maxima, minima, saddle points, and degenerate cases, as illustrated in Figure 1.11(b).
Two examples are given in Figure 1.12.
16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book
Volume Rendering 15
1.8 Functions on surfaces
Functions on surface visualization deals with the visual display of scalar functions
whose domain is restricted to an arbitrary geometric surface in three dimensions. The
surface may be the isosurface of another scalar eld, or simply a geometric domain
with an associated function eld [FLN+ 90, BOP92, BX94, BBX95].

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

16 Bajaj

Figure 1.13 Visualiztion of functions on surfacea. (top-left) The electrostatic energy poten-
tial shown on an isosurface of van der Waals interaction potenial energy. (top-right) Pressure
distribution around the earth globe. (bottom-left) Pressure distribution on the surface of a
jet engine modeled and displayed using tensor product surface splines. (bottom-right) Stress
distribution on a human knee joint based on static loads.

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

2

Ecient Techniques for

Volume Rendering of Scalar
Fields

DATA VISUALIZATION TECHNIQUES, Edited by C. Bajaj

c 1998 John Wiley & Sons Ltd
18 Bajaj

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

3

Accelerated IsoContouring of
Scalar Fields

DATA VISUALIZATION TECHNIQUES, Edited by C. Bajaj

c 1998 John Wiley & Sons Ltd
20 Bajaj

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

4

Surface Interrogation:
Visualization Techniques for
Surface Analysis

DATA VISUALIZATION TECHNIQUES, Edited by C. Bajaj

c 1998 John Wiley & Sons Ltd
22 Bajaj

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

5

Vector Field Visualization

Techniques

DATA VISUALIZATION TECHNIQUES, Edited by C. Bajaj

c 1998 John Wiley & Sons Ltd
24 Bajaj

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

6

Applications of Texture
Mapping to Volume and Flow
Visualization

DATA VISUALIZATION TECHNIQUES, Edited by C. Bajaj

c 1998 John Wiley & Sons Ltd
26 Bajaj

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

7

Continuous Bayesian Tissue

Classication for Visualization

DATA VISUALIZATION TECHNIQUES, Edited by C. Bajaj

c 1998 John Wiley & Sons Ltd
28 REFERENCES

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

References
[Baj88] C. Bajaj. Geometric modeling with algebraic surfaces. In D. Handscomb, editor,
The Mathematics of Surfaces III, pages 3{48. Oxford Univ. Press, 1988.
[BBB+ 97] J. Bloomenthal, C. L. Bajaj, J. Blinn, M.-P. Cani-Gascuel, A. Rockwood,
B. Wyvill, and G. Wyvill. Introduction to Implicit Surfaces. Morgan Kaufman
Publishers, 1997.
[BBX95] C. L. Bajaj, F. Bernardini, and G. Xu. Automatic reconstruction of surfaces and
scalar elds from 3D scans. In R. Cook, editor, SIGGRAPH 95 Conference Pro-
ceedings, Annual Conference Series, pages 109{118. ACM SIGGRAPH, Addison
Wesley, August 1995. held in Los Angeles, California, 06-11 August 1995.
[BCL96a] C. L. Bajaj, E. J. Coyle, and K.-N. Lin. Surface and 3D triangular meshes from
planar cross sections. In Proc. of the 5th International Meshing Roundtable,
number SAND96-2301 UC-405 in Sandia Report, pages 169{178, 1996.
[BCL96b] C. L. Bajaj, E. J. Coyle, and K.-N. Lin. Arbitrary topology shape reconstruction
from planar cross sections. Graphical Models and Image Processing, 6(58):524{
543, November 1996.
[BLMP97] C. Bajaj, H. Y. Lee, R. Merkert, and V. Pascucci. NURBS based B-rep models
for macromolecules and their properties. In C. Homann and W. Bronsvort,
editors, Proceedings of the 4th Symposium on Solid Modeling and Applications,
pages 217{228, New York, May14{16 1997. ACM Press.
[BOP92] R. Barnhill, K. Opitz, and H. Pottmann. Fat surfaces: a trivariate approach
to triangle-based interpolation on surfaces. Computer Aided Geometric Design,
9:365{378, 1992.
[Bow95] J. E. Bowie, editor. Data Visualization in Molecular Science. Addison-Wesley
Publishing Company, 1995.
[BPS97] C. L. Bajaj, V. Pascucci, and D. R. Schikore. The contour spectrum. In Pro-
ceedings of IEEE Visualization '97, pages 167{173, October 1997.
[BR94] C. Bajaj and A. Royappa. Triangulation and Display of Arbitrary Rational Para-
metric Surfaces. In R. Bergeron, A. Kaufman, editor, Proc. of IEEE Visualization
'94 Conference, 1994.
[BS95] C. L. Bajaj and D. R. Schikore. Decimation of 2d scalar data with error control,
1995.
[BS96a] C. L. Bajaj and D. R. Schikore. Error-bounded reduction of triangle meshes with
multivariate data. In Proceedings of SPIE Symposium on Visual Data Exploration
and Analysis III, 1996.
[BS96b] C. L. Bajaj and D. R. Schikore. Visualization of scalar topology for structural
enhancement, 1996.
[BS97] . Topology preserving data simplication with error bounds, 1997.
[BX94] C. Bajaj and G. Xu. Modeling Scattered Function Data on Curved Surface.
In J. Chen, N. Thalmann, Z. Tang, and D. Thalmann, editor, Fundamentals of
Computer Graphics, pages 19 { 29, Beijing, China, 1994.
[CDM+ 94] P. Cignoni, L. De Floriani, C. Montoni, E. Puppo, and R. Scopigno. Multiresolu-
tion modeling and visualization of volume data based on simplicial complexes. In
A. Kaufman and W. Krueger, editors, 1994 Symposium on Volume Visualization,

DATA VISUALIZATION TECHNIQUES, Edited by C. Bajaj

c 1998 John Wiley & Sons Ltd
30 REFERENCES

pages 19{26. ACM SIGGRAPH, October 1994. ISBN 0-89791-741-3.

[Cle93] W. S. Cleveland. Visualizing Data. Hobart Press, Summit, New Jersey, 1993.
[COL96] D. Cohen-Or and Y. Levanoni. Temporal continuity in levels of detail. In R. Yagel
and G. M. Nielson, editors, Visualization '96 Proceedings, pages 37{42, 1996.
[CPD+ 96] A. Certain, J. Popovic, T. DeRose, T. Duchamp, D. Salesin, and W. Stuetzle. In-
teractive multiresolution surface viewing. In H. Rushmeier, editor, SIGGRAPH
96 Conference Proceedings, Annual Conference Series, pages 91{98. ACM SIG-
GRAPH, Addison Wesley, August 1996. held in New Orleans, Louisiana, 04-09
August 1996.
[CVM+ 96] J. Cohen, A. Varshney, D. Manocha, G. Turk, H. Weber, P. Agarwal, Frederick
P. Brooks, Jr., and W. Wright. Simplication envelopes. In H. Rushmeier, editor,
SIGGRAPH '96 Conference Proceedings, Annual Conference Series, pages 119{
128, 1996. held in New Orleans, LA, August 4-9, 1996.
[Dau92] I. Daubechies. Ten Lectures on Wavelets, volume 61 of CBMS-NSF Regional
Conference Series in Applied Mathematics. Society for Industrial and Applied
Mathematics, Philadelphia, 1992.
[dBD95] M. de Berg and K. Dobrindt. On levels of detail in terrains. In Proc. 11th Annu.
ACM Sympos. Comput. Geom., pages C26{C27, 1995.
[DE95] Delnado and Edelsbrunner. An incremental algorithm for betti numbers of
simplicial complexes on the 3-sphere. Computer Aided Geometric Design, 12,
1995.
[Del94] T. Delmarcelle. The Visualization of Second-Order Tensor Fields. PhD thesis,
Stanford University, 1994. Ginzton Lab Report 5228.
[DFP85] L. DeFloriani, B. Falcidieno, and C. Pienovi. Delaunay-based representation of
surfaces dened over arbitrarily shaped domains. Computer Vision, Graphics
and Image Processing, 32:127{140, 1985.
[DJL92a] R. DeVore, B. Jawerth, and B. J. Lucier. Image compression through wavelet
transform coding. IEEE Transactions on Information Theory, 38:719{746, 1992.
[DJL92b] R. A. DeVore, B. Jawerth, and B. J. Lucier. Surface compression. Computer
Aided Geometric Design, 9:219{239, 1992.
[EDD+ 95] M. Eck, T. DeRose, T. Duchamp, H. Hoppe, M. Lounsbery, and W. Stuetzle.
Multiresolution analysis of arbitrary meshes. In R. Cook, editor, SIGGRAPH 95
Conference Proceedings, Annual Conference Series, pages 173{182. ACM SIG-
GRAPH, Addison Wesley, August 1995. held in Los Angeles, California, 06-11
August 1995.
[Ede95] Edelsbrunner. The union of balls and its dual shape. GEOMETRY: Discrete &
Computational Geometry, 13, 1995.
[FFNP84] L. D. Floriani, B. Falcidieno, G. Nagy, and C. Pienovi. A hierarchical structure
for surface approximation. Computers and Graphics, 8(2):183{193, 1984.
[FL79] R. J. Fowler and J. J. Little. Automatic extraction of irregular network digital
terrain models. In Computer Graphics (SIGGRAPH '79 Proceedings), volume
13(3), pages 199{207, 1979.
[FLN+ 90] T. Foley, D. Lane, G. Nielson, R. Franke, and H. Hagen. Interpolation of scattered
data on closed surfaces. Computer Aided Geometric Design, 7:303{312, 1990.
[FS93] T. A. Funkhouser and C. H. Sequin. Adaptive display algorithm for interactive
frame rates during visualization of complex virtual environments. In Computer
Graphics (SIGGRAPH '93 Proceedings), volume 27, pages 247{254, 1993.
[FS94] A. Finkelstein and D. H. Salesin. Multiresolution curves. In A. Glassner, edi-
tor, Proceedings of SIGGRAPH '94 (Orlando, Florida, July 24{29, 1994), Com-
puter Graphics Proceedings, Annual Conference Series, pages 261{268. ACM
SIGGRAPH, ACM Press, July 1994. ISBN 0-89791-667-0.
[GLL91] A. Globus, C. Levit, and T. Lasinski. A tool for visualizing the topology of
three-dimensional vector elds. In G. M. Nielson and L. Rosenblum, editors,
Visualization '91 Proceedings, pages 33{40, 1991.
[Ham94] B. Hamann. A data reduction scheme for triangulated surfaces. In Computer
Aided Geometric Design, volume 13, pages 197{214. 1994.

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

REFERENCES 31
[HDD+ 93] H. Hoppe, T. DeRose, T. Duchamp, J. McDonald, and W. Stuetzle. Mesh opti-
mization. In J. T. Kajiya, editor, Computer Graphics (SIGGRAPH '93 Proceed-
ings), volume 27, pages 19{26, 1993.
[HH89] J. Helman and L. Hesselink. Representation and display of vector eld topology
in uid ow data sets. IEEE Computer, 12(8):27{36, 1989.
[HH90] J. L. Helman and L. Hesselink. Surface representations of two- and three-
dimensional uid ow topology. In Proceedings of Visualization '90, pages 6{13,
1990.
[HHK+95] T. He, L. Hong, A. E. Kaufman, A. Varshney, and S. Wang. Voxel based ob-
ject simplication. In G. M. Nielson and D. Silver, editors, Visualization '95
Proceedings, pages 296{303, 1995.
[Hop96] H. Hoppe. Progressive meshes. In H. Rushmeier, editor, SIGGRAPH '96 Con-
ference Proceedings, Annual Conference Series, pages 99{108, 1996. held in New
Orleans, Louisiana, August 4-9, 1996.
[HU94] H. M. Hearnshaw and D. J. Unwin, editors. Visualization in Graphical Informa-
tion Systems. John Wiley & Sons, 1994.
[KK93] P. R. Keller and M. M. Keller. Visual Cues: Practical Data Visualization. IEEE
Computer Society Press and IEEE Press, 1993.
[KM92] D. Kenwright and G. Mallinson. A 3-D streamline tracking algorithm using dual
stream functions. In A. E. Kaufman and G. Nielson, editors, Proceedings of IEEE
Visualization '92 (Boston, Massachusetts, October 19{23, 1992), Visualization
Conference Series, pages 62{68, 1992.
[LC87] W. E. Lorensen and H. E. Cline. Marching cubes: A high resolution 3D surface
construction algorithm. In M. C. Stone, editor, Computer Graphics (SIGGRAPH
'87 Proceedings), volume 21, pages 163{169, 1987.
[Lee91] J. Lee. Comparison of existing methods for building triangular irregular networks.
Int. Journal of Geographical Information Systems, 5(2):267{285, 1991.
[Lev92] M. Levoy. Volume rendering using the Fourier projection-slice theorem. In
Proceedings of Graphics Interface '92, pages 61{69, 1992.
[LKR+ 96] P. Lindstrom, D. Koller, W. Ribarsky, L. F. Hodges, N. Faust, and G. A. Turner.
Real-time, continuous level of detail rendering of height elds. In H. Rushmeier,
editor, SIGGRAPH '96 Conference Proceedings, Annual Conference Series, pages
109{118, 1996. Held in New Orleans, Louisiana, August 4-9, 1996.
[Mal89] S. G. Mallat. A theory for multiresolution signal decomposition: The wavelet rep-
resentation. IEEE Transactions on Pattern Analysis and Machine Intelligence,
11(7):674{693, July 1989.
[Mal93] T. Malzbender. Fourier volume rendering. ACM Transactions on Graphics,
12(3):233{250, 1993.
[Mey94] D. Meyers. Multiresolution tiling. In Proceedings of Graphics Interface '94, pages
25{32, Ban, Alberta, Canada, 1994.
[Mur92] S. Muraki. Approximation and rendering of volume data using wavelet trans-
forms. In A. E. Kaufman and G. M. Nielson, editors, Proceedings of Visualization
'92 (Boston, Massachusetts, October 19{23, 1992), pages 21{28. IEEE Computer
Society, IEEE Computer Society Press, October 1992.
[Mur95] S. Muraki. Multiscale volume representation by a DoG wavelet. IEEE Transac-
tions on Visualization and Computer Graphics, 1(2):109{116, June 1995. ISSN
1077-2626.
[NHM97] G. M. Nielson, H. Hagen, and H. Muller, editors. Scientic Visualization. IEEE
Computer Society, 1997.
[PDDT94] E. Puppo, L. Davis, D. DeMenthon, and Y. Teng. Parallel terrain triangulation.
Int. Journal of Geographical Information Systems, 8(2):105{128, 1994.
[Pup96] E. Puppo. Variable resolution terrain surfaces. In F. Fiala, E. Kranakis, and J.-R.
Sack, editors, Canadian Conference on Computational Geometry, pages 202{210,
1996.
[RB93] J. Rossignac and P. Borrel. Multi-resolution 3D approximations for rendering
complex scenes. In IFIP TC 5.WG 5.10 II Conference on Geometric Modeling

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book

32 REFERENCES

in Computer Graphics, 1993. IBM Research Report RC 17697, Yorktown Heights,

NY 10598.
[REE+ 94] L. Rosenblum, R. A. Earnshaw, J. Encarnacao, H. Hagen, A. Kaufman, S. Kli-
menko, G. Nielson, F. Post, and D. Thalmann, editors. ScienticVisualization :
Advances and Challenges. Academic Press, 1994.
[RR96] R. Ronfard and J. Rossignac. Full-range approximation of triangulated polyhe-
dra. In Eurographics '96, volume 15. 1996.
[SDS96] E. J. Stollnitz, T. D. DeRose, and D. H. Salesin. Wavelets for computer graphics:
Theory and applications, 1996.
[SMK95] C. Silva, J. Mitchell, and A. E. Kaufman. Automatic generation of triangular
irregular networks using greedy cuts. In G. M. Nielson and D. Silver, editors,
Visualization '95 Proceedings, pages 201{208, 1995.
[SSO96] M. Sanner, J.-C. Spehner, and A. Olson. Reduced surface: an ecient way to
compute molecular surfaces. Biopolymers, 38(3):305{320, 1996.
[SZL92] W. J. Schroeder, J. A. Zarge, and W. E. Lorensen. Decimation of triangle meshes.
In E. E. Catmull, editor, Computer Graphics (SIGGRAPH '92 Proceedings),
volume 26(2), pages 65{70, 1992.
[TL93] T. Totsuka and M. Levoy. Frequency domain volume rendering. In J. T. Kajiya,
editor, Computer Graphics (SIGGRAPH '93 Proceedings), volume 27, pages 271{
278, August 1993.
[Tsa93] V. J. D. Tsai. Delaunay triangulations in TIN creation: an overview and a linear-
time algorithm. Int. Journal of Geographical Information Systems, 7(6):501{524,
1993.
[Tur92] G. Turk. Re-tiling polygonal surfaces. In E. E. Catmull, editor, Computer Graph-
ics (SIGGRAPH '92 Proceedings), volume 26(2), pages 55{64, 1992.
[Wat92] D. F. Watson. Contouring. Pergamon, 1992.
[WJ92] J. M. Ware and C. B. Jones. A multiresolution topographic surface database.
Int. Journal of Geographical Information Systems, 6(6):479{496, 1992.
[WV94] J. Wilhelms and A. Van Gelder. Multi-dimensional trees for controlled volume
rendering and compression. In A. Kaufman and W. Krueger, editors, 1994 Sym-
posium on Volume Visualization, pages 27{34. ACM SIGGRAPH, October 1994.
ISBN 0-89791-741-3.
[ZSS97] D. Zorin, P. Schroder, and W. Sweldens. Interactive multiresolution mesh editing.
In T. Whitted, editor, SIGGRAPH 97 Conference Proceedings, Annual Confer-
ence Series, pages 259{268. ACM SIGGRAPH, Addison Wesley, August 1997.
ISBN 0-89791-896-7.

16/4/1998 12:20|PAGE PROOFS for John Wiley & Sons Ltd|book