UNIT-3
Unit-3:
Visualization Techniques: Spatial Data: One-Dimensional Data - Two-Dimensional Data –
Three-Dimensional Data - Dynamic Data - Combining Techniques.
References:
1. Matthew Ward, Georges Grinstein and Daniel Keim, : Interactive Data Visualization
Foundations, Techniques, Applications “,2010
3.1 Visualization Techniques for spatial data:
Spatial data visualization, which corresponds to the field of scientific visualization,
assumes that the data has an implicit or explicit spatial or spatio temporal attribute.
This constraint facilitates both the creation and interpretation of the visualization, as there is
an intuitive, and often straightforward, mapping of the data attributes to graphical attributes of
the entities conveying the information in the visualization.
As our visual perception system is constantly receiving and interpreting imagery of
the physical phenomena surrounding us, it is quite natural to process the same form of
imagery on a computer screen.
In creating a visualization of spatial data, we must decide what spatial attributes of the data
will map to the spatial attributes (locations) on the screen.
This can involve many forms of transformation, including scaling, rotation, translation,
shearing, and projection.
Once the spatial attributes are accommodated, other attributes and values associated with
the data must then be mapped to other components of the visualization, whether it is an
attribute such as color or texture or the size or shape of a graphical entity.
In creating a visualization of spatial data, we must decide what spatial attributes of the data
will map to the spatial attributes (locations) on the screen. This can involve many forms of
transformation, including scaling, rotation, translation, shearing, and projection. Once the
spatial attributes are accommodated, other attributes and values associated with the data
must then be mapped to other components of the visualization, whether it is an attribute
such as color or texture or the size or shape of a graphical entity.
One-Dimensional Data:
One-dimensional spatial data is often the result of accumulating samples or readings of
some phenomenon while moving along a path in space. For example, a drill-hole sample
will contain mineral content and ore grade information based on the distance from the top of
the drill-hole. This sort of sampling is often referred to as a probe when exploring structures
of higher dimensions.
Given a one-dimensional sequence of univariate data (only one value per data item), we can
map the spatial data to one of the screen dimensions and the data value itself to either the
other screen dimension (to form a line graph, see Figure ) or to the color of a mark or region
along the spatial axis (to form a color bar). The data needs to be scaled to fit within the range
of the display attribute (either number of pixels or number of colors). Parts of the display
space or color range might be reserved for other aspects of the visualization, such as the axes,
labels, and key, so the most general structure for an algorithm to generate such a
visualization will use parameters for the bounds of both the data and display spaces.
Fig : A line graph of a 1D sequence of data values
Assume that (datamin, datamax) are computed as the minimum and maximum values for the
data, and datacount indicates the number of data points to be displayed (this could be all of
the data, or just a selected subset). Also assume the section of the display that will hold the
visualization is defined by the rectangle (xmin, ymin, xmax, ymax). To generate a line graph
from an array
called “data,” we could use the following code (assuming the drawing
command draw-Line exists).
Draw-Line-Graph(data, dataCount, xMin, xMax, yMin, yMax)
1 dataMin ← computeMin(data, dataCount)
2 dataMax ← computeMax(data, dataCount)
3 xFrom ← xMin
4 yFrom ← worldToScreenY(data[0], dataMin, dataMax, yMin,
5 yMax)
6 for i ← 1 to dataCount
7 do xTo ← worldToScreenX(i, dataCount, xMin, xMax)
8 yTo ← worldToScreenY(data[i], dataMin, dataMax,
9 yMin, yMax)
10 drawLine(xF rom, yFrom, xTo, yTo)
11 xFrom ← xTo
12 yFrom ← yTo
worldToScreenX(index, dataCount, xMin, xMax)
return (xMin + index ∗ (xMax − xMin)/dataCount)
worldToScreenY(value, dataMin, dataMax, yMin, yMax)
return (yMin + (value − dataMin) ∗ (yMax − yMin)/(dataMax − dataMin))
Note that each transformation is a combination of offsets and scaling that we use to convert
values in one coordinate system into another. This operation is used heavily in computer
graphics, and we will use variants on it numerous times in this book. It is derived based on
the observation that a point within a given range can be transformed to a point within a
different range by computing the relative position of the point within the range, e.g., what
percentage of the distance from one end of the range to the other is the selected point.
Mathematically, this is given as
(Ai − Amin)/(Amax − Amin)=(Bi − Bmin)/(Bmax − Bmin)
where Ai is a location in the range [Amin → Amax] and Bi is the corresponding location in
the range [Bmin → Bmax].
Two-Dimensional Data:
Data with two spatial dimensions get visualized predominantly by mapping the spatial
attributes of the data to the spatial attributes of the screen. The result can be one of the
following visualizations:
1. An image results if a single data value at each location is mapped to color and
all intermediate pixels are colored via interpolation .
2. A rubber sheet results if the data, whether regularly or irregularly spaced, is mapped
to the height of a point in three dimensions, with the points triangulated so that a
surface can be formed.
3. A cityscape is formed by drawing three-dimensional objects (generally boxes) at
locations on a plane, where the data can control the attributes of the graphical objects
(i.e., height and color).
4. A scatterplot results if, at each location on the plot, the data value(s) control the
color, shape, or size of a marker. Note that unlike for images, no interpolation is
performed.
5. A map results if the data contains linear and area features, as well as point objects. A
linear feature, such as a road or stream, is represented as a sequence of connected
coordinates, which are plotted as a series of line segments. Area features, such as a
lake or political boundary, are generally represented as a closed contour, a set of
coordinates where the first and last points are the same.
6. A contour or isovalue map conveys boundary information extracted from an image
depicting a continuous phenomenon, such as elevation or temperature. The term
isovalue means “single value,” and thus a contour on such a map indicates the
boundary between points above this value and points below the value. It can be
formed by considering two-by-two arrays of adjacent data values as the corners of a
rectangle or square, and generating edges across this rectangle when one or more
values are on the opposite side of the isovalue from one or more of the others.
Probing Two-Dimensional Data
Besides the visualization techniques that display the entire data set, we can also
choose to visualize one-dimensional subsets, projections, or summarizations of the
data. After reducing the dimensionality in some manner, we are then free to employ
any of the techniques described in the previous section. Let us examine some of the
projection options in detail.
Frequency histograms. We can summarize data sets of arbitrary dimensional ity by
computing the frequency at which values or subranges of values occur in the data, and
then display this information as a bar graph. Important considerations when using
subranges are how to decide the number of sub ranges to use, and where the breaks
between subranges should be. Features of importance in the data might be lost if this
is done incorrectly, and it is best performed by someone familiar with the data
characteristics. Sim ply dividing the range into a fixed number of evenly sized
subranges, while simple, is rarely an effective strategy.
Row and column aggregations. Visually depicting summarizations of the rows
and/or columns of an image is a useful mechanism for localizing the bound aries of
features of interest and regions of low or high variability. Many de scriptors can be
used for this purpose, including the sum, average, median, standard deviation,
maximum, or minimum values. The resulting one dimensional visualization can be
viewed separately or, as often the case, placed alongside the two-dimensional
visualization as supplemental informa tion. Color bars, line plots, and bar graphs have
all been used with success in this type of visualization.
Linear probes. A one-dimensional probe of a two-dimensional data set can be
likened to a drill hole for mineral exploration. A line is passed through the data, and
the values encountered are visualized using one of the one dimensional techniques
previously described. To accomplish this, we use two mathematical tools already
presented in this chapter: parametric equations and bilinear interpolation. We start by
creating a parametric equation of our probe using input from the user (either a pair of
points or a point and a direction). If we assume that this line was specified as two sets
of coordinates, P1 and P2, the parametric equation of the line segment joining the
points is simply P(t)=P1 + t(P2 − P1), where 0 ≤ t ≤ 1.0. We can then get the
coordinate of an arbitrary point along this line segment by choosing a value of t
within this range. Generally, these are evenly spaced, and the number of samples
computed is proportional to the length of the line segment (so that probes that barely
clip a corner of the data grid do not get represented by the same number of samples as
a probe along the diagonal of the grid). Given the coordinates of the sampled points,
we can now use the interpolation procedure to compute values, which can then be
visualized as a one-dimensional data set
.Three-Dimensional data:
As with two-dimensional data, three-dimensional spatial data may be either discrete
samples of a continuous phenomenon or a structure best described via vertices, edges,
and polygons. In reality, many visualizations of science and engineering data contain a
combination of these data representations, such as air flow around a wing or stress attributes
of a mechanical part.
1 .Visualizing Explicit Surfaces:
An explicit surface is one that has been defined in one of two ways:
1. a list of three-dimensional vertices, a list of edges, i.e., connections between the
vertices (specified as a pair of indices into the vertex list), and a list of planar
polygon patches (usually specified as a fixed or variable length list of indices into
the edge list);
2. a set of parametric equations for defining the x-, y-, and z-coordinates of points on the
surface, along with an interconnection strategy (e.g., a triangular or rectilinear grid)
for computing the edges and patches. The step size of the parameters can be used to
control the smoothness of the curved surface.
2. Visualizing Volume Data:
As pixels are to two-dimensional visualization, voxels, or volume elements, are to
three- dimensional visualization. Volume data is generally a sampling of a continuous
phenomenon, and can be either acquired via sensors (e.g. tomographic data sets) or
generated via simulations (e.g., computational fluid dynamics). In each case, we have
one or more data dimensions with regular or irregular positions, and the goal is to
convey to the viewer the structure, patterns, and anomalies within the data.
Most approaches to visualizing volume data fall into one of the following
categories :
Slicing techniques. Using a cut plane, either aligned with an axis or arbitrarily
oriented, probe the data to extract a two-dimensional slice of the data, and then use
one of the two-dimensional spatial data visualization methods.
Isosurface techniques. Given a user-specified value, generate a surface description
and visualize it using one of the explicit surface visualization
techniques.
Direct volume rendering. Either cast rays into the volume and compute a pixel
value based on the data encountered by the ray, or project each voxel onto the
projection plane using some method of accumulating effects on pixels.
Direct Volume Visualization Techniques. Direct volume rendering means that no
three-dimensional polygons are created for use with traditional graphics rendering
techniques. Rather, pixels in the resulting image are computed on an individual basis,
either by casting rays through the pixel through the volume, or by projecting voxels
onto the plane of projection.
The basic process of rendering a volumetric data set starts by transforming the
positions of the voxels into the viewing coordinate system, following the same
procedure used in the traditional three- dimensional graphics pipeline. This may or
may not include perspective distortion. The viewer must specify a view reference
point (the origin of the plane of projection), a view direction (the normal to the plane
of projection), the height and width of the image to be projected on the plane of
projection, and, for a perspective projection, the distance from the camera to the
plane of projection. The reader is referred to textbooks on computer graphics for
details of this process.
Once the voxels have been positioned, we have the option of either:
forward mapping—project each voxel onto the plane of projection and determine
which pixels will be affected and in what way;
inverse mapping, also called ray casting—send a ray from each pixel in the plane of
projection through the volume, sampling values along the ray, and determining the
resulting value for each pixel.
Dynamic Data
Flow visualization is the study of methods to display the dynamic behavior in
liquids and gases. The field dates back at least to the mid-1400s, where Leonardo
Da Vinci sketched images of fine particles of sand and wood shavings
that had been dropped into flowing liquids. Since then, laboratory flow
visualization has become more and more exact, with careful control of the
particulate size and distribution. Advances in photography have also helped
extend our understanding of how fluids flow under various circumstances.
More recently, computational fluid dynamics (CFD) has extended the
abilities of scientists to study flow by creating simulations of dynamic behavior
of fluids under a wide range of conditions. The result of this analysis
is usually a 2D or 3D grid of velocity vectors, which may be uniformly or
nonuniformly spaced. The goal is then to analyze this vector field to identify
features such as turbulence, vortices, saddle points, and other forms of
structure.
There are several variations on the structure of the field data that is
generated by these experiments. A static field is one in which there is only
a single, unchanging velocity field. Time-varying fields may either have
fixed positions with changing vector values or both changing positions and
changing vectors (for example, when modeling rotating turbine blades or
pitching airfoils). These latter types are referred to as unsteady.
Combining Techniques:
Many effective visualizations are actually combinations of two or more of the
techniques . Each technique has its strengths and weaknesses in terms of the types of
information it can or cannot effectively visualize, so
a combined visualization, as long as occlusion is minimized, can generate
results that are the sum of the strengths. At the same time, more and more
problems require the simultaneous analysis of multiple data sets to arrive at an
informed result.
A. Slice Plus Isosurface:
The isosurface is capable of conveying surface structure, which is difficult to
obtain from volume slicing, even with animation of the slice position.
However, the isosurface only provides information on a single value within
the entire volume, with no indication of the distribution of other values or the
gradient (rate of change) of the selected value at different locations.
The slice provides very detailed two-dimensional information, especially with
an appropriate choice of color assignments. It can convey to the user the
regions of relative uniformity, as well as those exhibiting significant change.
Another advantage is that the image slice can convey nested regions of a
particular value range, while the isosurface, in general, will only display the
outer-most surface.
B. Isosurface Plus Glyphs:
isosurfaces are useful for conveying details of three-dimensional surfaces, but,
in general, do not incorporate other aspects of the data. Glyphs , such as the
popular arrow glyph, can be used to display the magnitude and direction of
change within a data set, either as a gradient in static data or a flow in dynamic
data. The glyphs may be positioned in close proximity to the isosurface, since
these are known to be positions of interest, or the positions may be controlled
separately
C. Rubber Sheet plus Contour Lines and Color:
In this example, we start with a rubber sheet to convey a two- dimensional
field of values as a height field. This can reveal peaks and valleys found in the
data, and by creating a virtual landscape, we can
build on the users’ intuition about interpreting the data in a topographic
manner . We can augment this visualization by mapping color to the elevation,
thus making it easier to identify widely separated regions of similar height .
Finally, we can superimpose contour lines at certain levels, making the
gradient information much more apparent.