Overview and State-of-the-Art of Uncertainty Visualization - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 21

1.7 Open Problems
1.7.1 Perceptual and Cognitive Implications
Since visualization often relies heavily on the use of colors to convey information, it
can be quite challenging for individuals with color vision deficiency. For them, even
interpreting visualizations that would pose no problems for individuals with normal
color vision can be a difficult task. In this case, however, the resulting ambiguity, and
therefore, uncertainty, is inherent to the observer, falling outside the broad sources
of uncertainty discussed in Sect. 1.1.1 (i.e., uncertainty observed in sampled data,
uncertaintymeasures generated bymodels or simulations, and uncertainty introduced
by the data processing or visualization processes). Thus, individuals with color vision
deficiency have to constantly deal with uncertainty visualizations and make decisions
based on ambiguous information. For those individuals, the display of additional data
that tries to express the amount of uncertainty fromvarious sourcesmay even generate
further ambiguities. The issues involving uncertainty visualization and color vision
deficiency are discussed in Chap. 2 .
1.7.2 Comparative Visualizations
The visualization of uncertainty may involve a comparison of different results, such
as aweather forecast generatedwith different parameters. To detect similarities or dif-
ferences in the results a comparative visualization technique [ 73 ] can be employed.
In 3D a visualization via fusion [ 10 , 12 ] is not feasible beyond a small number
(2 or 3) of data sets, due to clutter and inter-dependence of the different data sets. An
alternative to fusion is a side-by-side view of the data sets. This may be problematic
in 3D since it is hard to find corresponding reference points in more than two vol-
umes. As an example to control a 3D comparison Balabanian et al. [ 1 ] propose to
integrate volume visualization into a hierarchical graph structure. These integrated
views provide an interactive side-by-side display of different volumes while the
parameter space can be explored through the graph structure. In 2D a blending of
different results has basically the same issues as a fusion in 3D [ 31 , 35 ]. There are
techniques which allow a comparative visualization of different data sets in a single
image. Urness et al. [ 104 ] introduced color weaving for flow visualization to com-
pare different flow fields in a single 2D view. In contrast to blending, each pixel of
the resulting image represents an unmodified value from one of the data sets. The
generated pattern provides a good overview to detect similar or varying regions in
the data sets. To compare certain regions in more detail, e.g., borders, it is better to
consider larger comparison areas than individual pixels. In this context it is crucial
that data sets which should be compared are visualized next to each other to get
a direct comparison for a certain area. For only two data sets a checkerboard pat-
tern can be used to achieve screen door transparency [ 95 ]. The white squares show