Visual Exploration of Multivariate Volume Data Based on Clustering - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 174

ilar behavior. This aggregation step allows for interactive visual analyses of the
data. While purely interactive methods have limits when trying to operate in a
multi-dimensional attribute space, aggregation via clustering allows for new visual
encodings of the attribute space. Different visual encodings of the clustering result
provide different information such that a system with different coordinated views
on the clustering result proves to be most efficient. Those coordinated views shall
also allow for interactive modification of the clustering result during the exploration
process, especially if the user can bring in some additional domain expertise.
The presented approaches mainly deal with multivariate data in the sense of
multiple scalar fields, where the scalar fields may be given dimensions or may be
derived attributes. In particular, the attributes may be derived from vector or tensor
fields. The most obvious future direction is to extend the approaches to time-varying
data. Akiba and Ma [ 1 ] did incorporate the time dimension into their interactive
system, however, it was not based on the clustering idea. Still, the ideas may be
applicable to cluster-based approaches. Next, it is of interest to also incorporate spa-
tial information into the clustering approach such that one cluster always represents
a connected region in object space. Local spatial relationships have been used, e.g.,
to draw statistics [ 11 ], which allows for further differentiation between regions of
similar values but different textures. However, these local measures do not guarantee
any global properties. In terms of the clustering itself, scaling and normalization
issues of the individual dimensions may need further investigation, especially when
simultaneously considering original attributes, derived attributes, spatial dimensions,
time dimension, etc. Non-linear scalings may be appropriate to use. Finally, it has
not yet been investigated how these approaches generalize to multi-run or ensemble
data, although a cluster-based visualization of the parameter space is also feasible.
References
1. Akiba, H., Ma, K.L.: A tri-space visualization interface for analyzing time-varying multivariate
volume data. In: Proceedings of Eurographics/IEEE VGTC Symposium on Visualization, pp.
115-122, May 2007
2. Akiba, H., Ma, K.L: A tri-space visualization interface for analyzing time-varying multivariate
volume data. In: EuroVis07—Eurographics/IEEE VGTC Symposium on Visualization, pp.
115-122, May 2007
3. Akiba, H., Ma, K.L., Chen, J.H., Hawkes, E.R.: Visualizing multivariate volume data from
turbulent combustion simulations. Comput. Sci. Engg. 9 (2), 76-83 (2007)
4. Andreas K.ö.: A survey of methods for multivariate data projection, visualisation and inter-
active analysis. In: Yamakawa, T. (ed.) 5th International Conference on Soft Computing and
Information/Intelligent Systems, pp. 55-59, Iizuka, Japan, Oct 16-20 (1998)
5. Bachthaler, S., Weiskopf, D.: Continuous scatterplots. IEEE Trans. Vis. Comput. Graph. (Pro-
ceedings Visualization/Information Visualization 2008), 14 (6), 1428-1435, Nov-Dec 2008
6. Bachthaler, S., Weiskopf, D.: Efficient and adaptive rendering of 2-d continuous scatterplots.
vol. 28, pp. 743-750 (2009)
7. Bezdek, J.C.: Pattern recognition with fuzzy objective function algorithms. Plenum Press, New
York (1981)