Overview and State-of-the-Art of Uncertainty Visualization - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 15

Case studies of specific data have been performed by Kao et al. [ 46 , 47 ]. Their
data sets come from NASAs Earth Observing System (EOS) Satellite images and
Light Detection And Ranging (LIDAR) data. The methods used to show this data
include encoding the mean as a 2D color map, and using standard deviation as a
displacement value. Histograms are also employed to understand better the density
of the PDFs. To explore the mode of specific distributions, a small set of PDFs are
plotted onto a color mapped spatial surface.
1.5.2 Uncertainty Visualization
Many visualization techniques that incorporate uncertainty information treat uncer-
tainty like an unknown or fuzzy quantity; [ 75 ] is a survey of such techniques. These
methods employ the meaning of the word uncertainty to create the interpretation
of uncertainty or unknown to indicate areas in a visualization with less confidence,
greater error, or high variation. Ironically, while blurring or fuzzing a visualization
accurately indicates the lowered confidence in that data, it does not lead to more
informed decision making. On the contrary, it obfuscates the information that leads
to the measure of uncertainty. Because it obscures rather than elucidates the quanti-
tative measures leading to the uncertain classification, such a solution to the problem
of adding qualitative information to visualization misses important information.
1.5.2.1 Comparison Techniques
Often, uncertainty describes a comparison that can most clearly be understood visu-
ally, such as the difference between surfaces generated using different techniques, or
a range of values that a surface might fall in. A simple approach to the visualization of
this type of information is a side-by-side comparison of data sets. An example of this
type of visualization is presented in Jiao et al. [ 41 ] where streamlines computed from
various fiber tracking algorithms are interactively displayed along with the global
and local difference measures. Another example is the time window, presented in
[ 112 ], in which temporal uncertainty around archeological sites is displayed, using
various visual clues, in an interactive, exploratory system.
However, this approach may not clearly manifest subtle differences when
the data are nearly the same, and it becomes harder to perform this comparison as the
visualization becomes more complicated. Another simple approach is to overlay the
data to be compared [ 45 ]. With this technique, the addition of transparency or wire
frame can produce a concise, direct comparison of the data sets. A similar approach
uses difference images to display areas of variation [ 108 ]. These approaches are
less effective, however, when the uncertainty can be categorized as more of a range
of values rather than just two distinct ones. In such cases, a surface sweep, known
as a fat surface [ 75 ], can be used to indicate all possible values. Another approach
is the integration of isosurface and volume rendering. Here, an opaque isosurface