Feature-Based Visualization of Multifields - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 178

more often features, which are abstract objects, are visualized in combination with
more traditional techniques. Furthermore, features can be used to guide the placement
of visualization objects such as glyphs or streamline seeds [ 3 , 36 , 38 ]. Finally, in time-
dependent datasets, features can be tracked over time [ 29 ], providing information on
the dynamics of processes. Feature-based visualizationmethods have been developed
for scalar, vector, and tensor fields in a wide range of application areas such as fluid
flow [ 27 ] and medical visualization [ 4 ].
17.2 Multifield Feature Definitions
17.2.1 Single-Field Versus Multifield Features
Multifield features and their definitions can only be fully understood in the context of
single-field feature definitions. While multifield and single-field features share com-
mon properties, they often present researchers and users with additional challenges
with respect to data size, limitations of visual space, and conceptional differences as
highlighted in the following.
Feature definitions and visualization techniques in a single-field setting benefit
from a set of assumptions. Often features are free of spatial overlaps, have equal
prioritywith respect to the visual space they occupy if no feature strength ismeasured,
and, most importantly, the feature definition process is a one-to-many mapping from
data to feature space. Contrarily, the feature extraction and visualization process
for a multifield data set is inherently a many-to-many mapping, thus introducing an
additional dimension of complexity.Where a single-field feature extraction technique
may produce a set of features for different feature definitions on the same field that
are combined into a final visualization, features in a multifield data set can be created
by feature definitions based on a single-field, multiple heterogeneous fields, or, in
the most complex case, multiple heterogeneous fields.
In situations, where feature based visualization consists of visual blending of
single-field features, the feature extraction and visualization process is extremely
similar to that in a single-field setting. True combined feature extraction techniques
are unique to multifield data in such that the geometric representation of the final
feature is dependent on a set of single fields.
17.2.2 Classes of Multifield Feature Definitions
Feature definitions in a multifield setting may be classified into one of two groups:
(a) Isolated Feature Definitions : Feature definition and extraction is performed
on an independent per-field level. The multifield notion is obtained as the final
visualization combines these isolated features into a common representation.
The resulting multifield feature is essentially a collection of classic features.