Image Processing Reference
In-Depth Information
26.4 Discussion and Future Directions
While the state of the art in large-data integration-based visualization has consider-
ably matured recently, a satisfactory method that can satisfy all visualization require-
ments remains elusive. Future research must address a number of difficult problems:
While data sets of significant size have proven feasible, the approaching exascale
generation of hardware architectures will impose new demands on visualization
algorithms; this is especially the case for parallel integral curve algorithms. It is
unclear at this point whether the load balancing strategies discussed in Sect.
26.3.3
can be adapted to such scenarios. More general schemes are required that provide
adequate scalability and do not require data pre-analysis, which will likely be
prohibitive on exascale data.
Current algorithms still require minutes to achieve results even for small prob-
lems due to the factors discussed in Sect.
26.2.4
. This stands in stark contrast with
the requirements of user-guided, interactive exploration that has proven extremely
valuable in obtaining insight into the complex nature of vector fields. Here, pro-
gressive algorithms are needed that can quickly produce approximate results which
are refined over time. However, the non-local nature of integral curves has thus far
prevented that application of downsampling techniques on this problem.
Similarly, as the volume of data resulting from simulation codes will grow past
the point of feasible retention on external storage, it is anticipated that future visu-
alization algorithms will possess a large in situ component, requiring the majority
of the analysis to be performed during a simulation run. At this point, it is entirely
unclear how to achieve user-guided vector field visualization in an in situ scenario.
While first steps in these directions have been taken, many problems remain open,
and more research in this interesting area is needed.
References
1. Becker, B. G., Max, N. L., Lane, D. A.: Unsteady flow volumes. In: Proceedings of IEEE
Visualization, pp. 329-335 (1995)
2. Bruckschen, R., Kuester, F., Hamann, B., Joy, K. I.: Real-time out-of-core visualization of par-
ticle traces. In: Proceedings of the IEEE Symposium on Parallel and Large-Data Visualization
and Graphics (PVG), pp. 45-50, Piscataway, NJ, USA, IEEE Press (2001)
3. Camp, D., Childs, H., Chourasia, A., Garth, C., Joy, K. I.: Evaluating the benefits of an extended
memory hierarchy for parallel streamline algorithms. In: Proceedings of the IEEE Symposium
on Large-Scale Data Analysis and Visualization (LDAV), IEEE Press (2011)
4. Camp, D., Garth, C., Childs, H., Pugmire, D., Joy, K.I.: Streamline integration using MPI-
Hybrid parallelism on a large multicore architecture. IEEE Trans. Vis. Comput. Graph.
17
(11),
1702-1713 (2011)
5. Chen, L., Fujishiro, I.: Optimizing parallel performance of streamline visualization for large
distributed flow datasets. In: Proceedings of IEEE VGTC Pacific Visualization Symposium
2008, 87-94 (2008)
Search WWH ::
Custom Search