Feature-Based Visualization of Multifields - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 179

(b) Combined Feature Definitions : Feature definition and extraction is performed
on a combined set of fields, i.e., the feature definition depends on information
encoded in multiple fields. These feature definitions extend the classic single-
field feature definition and creates features that are not present in any single
field.
There is generally no preferred way of multifield feature extraction and visualiza-
tion, neither in theory nor in applications. Both feature definitions are widely used
in the visualization community.
17.3 Classification of Visualization Techniques
17.3.1 Isolated and Modulated Features
The earliest and most straight-forward feature-based visualization techniques that
rely on data from multiple fields make use of individual feature definitions for each
field and combine these isolated features into a combined multifield visualization.
Often, complex feature extraction is limited to one field only with other fields used
for modulation of the extracted feature representation.
One of the earliest example is the enhanced display of streamlines in a flow
field as streamtubes and streamribbons by modulating extracted line geometry with
properties of the derived rotation and divergence fields [ 34 ]. Similarly, integral lines in
time-varying flow fields have been enhanced with deformation information extracted
from the derived velocity gradient field [ 24 ]. Both examples illustrate how the initial
feature extraction process is limited to one field, whereas the final visualization
conveys multifield data.
Love et al. [ 19 ] demonstrate the use of streamline extraction for multivalue data
sets. While the feature extraction process is limited to an individual field, concurrent
visualization of streamlines extracted from multiple related fields facilitates visual
comparison of flow directions.
An approach that modulates the geometry of the shared computational mesh of
multifield CFD data is presented by Henze [ 14 ]. Henze proposes to map the initially
feature-less computational grid of the simulation domain into spaces of fields present
in the data set. A single vector valued field as well as multiple scalar valued fields are
used to displace vertices of the original mesh. Feature identification and extraction
in this new coordinate system allows joint analysis of the affected fields.
17.3.2 Locally Defined Features
One of the classical sources of multifields is Computational Fluid Dynamics (CFD)
simulations, the results of which contain physical variables such as velocity, pressure,
and density. CFD solvers commonly output an extensive set of scalar-, vector-, or