Contribution to the research
seminar on visualizatino in winter semester 1998/1999
by Helwig
Löffelmann, Institute
of Computer Graphics, Vienna University of Technology, Austria.
Goal [slide 2]: Visualization application should be interactive, i.e., respond in ``real-time''. This means that response time (time of updating visualization after user interaction) should not exceed 0.2-0.5 s [1].
Specifics [slide 3]: As visualization application we consider software that is (more or less) based on the visualization pipeline (data aquisition, filtering/etc., visualization mapping, rendering), i.e., follows a sequential process composed of various steps to convert data into visualization results. Several parallel pipelines can be imagined. User interaction is classified: (1) viewing interaction (changes to viewpoint, lightning conditions, etc., i.e., last step in the pipeline), (2) mapping interaction (changes to the mapping stage, and (3) source data interaction (modifications to the data on which the visualization is based [2]. Different kinds of user interaction cause different needs of recomputation: repeated rendering (case 1), recomputation of geometry (case 2), complete recomputation of entire pipeline (case 3).
Problem setup [slide 4]: The response time of the visualization application depends on different aspects: (A) what kind of user interaction is encountered, (B) the given parameters to involved pipeline steps (visualization parameters), and (C) the interdependencies between pipeline steps. Usually time constraints given by a certain pipeline module are proportional to the amount of input data. Thus, visualization parameters also (in addition to their direct influence) have indirect influence to modules further down the pipeline (see figure below).
Crucial question [slide 5]: Once a certain visualization setup is running, the crucial question is: how to change the setup, such that a too slow pipeline is accelerated, or a too course visualization process (does not usr all of the available time) is refined. Usually there is a huge number of possibilities to choose from, but the criterium is to head for optimal visualization results possible within the given constraints. This question kind of includes the question for the initial setup as well as the question what visualization modules to choose (if there are more than one to choose from).
Some of the possibles approaches to deal with this problem are:
| [1] | S. Bryson: Time management, simultaneity and timecritical computation in interactive unsteady visualization environments. Published in proceedings IEEE Visualization '96, 1996. |
| [2] | L. Mroz,
H. Löffelmann, and
E. Gröller:
Bringing Your Visualization Application to the
Internet. Published as technical report TR-186-2-98-14 at
the Institute of Computer Graphics, Vienna University of
Technology, 1998.
WWW-Links: technical report [.ps.gz], abstract. |