VISFILES

Vol.34 No.1 February 2000

---

ACM SIGGRAPH

Harnessing the Web for Scientific Visualization

Bill Hibbard
Space Science and Engineering Center
University of Wisconsin - Madison

February 00 Columns


Visfiles

I always learn something during Ken Brodlie’s presentations, so I have asked him to contribute this issue’s VisFiles column. His topic is the Web as a medium for visualization, which is certain to be increasingly important to all of us.

- Bill Hibbard

Ken Brodlie
Stuart Lovegrove
Jason Wood
University of Leeds

Introduction

Arguably the trigger for the success of the World Wide Web in the early 1990s was its ability to convey images as well as text. Documents could include colour pictures at no extra cost. In those days, however, scientific visualization in a Web context was a passive activity. A scientist would create a visualization using a conventional system, and publish it on a website - for a reader to later observe. The Web was simply a publication medium. Today of course the Web represents a rich distributed computing environment: visualization can now be executed as a live process with active engagement from the audience. A new field of Web-based visualization has emerged.

The Beginnings - The Threads of Web-based Visualization

Pioneering work on Web-based visualization was done by Ang et al [1]. They exploited the MIME-typing concept to allow visualization data to be sent over the Web and be processed by a Web browser. They linked their medical visualization system, VIS, to the Mosaic browser via its CCI interface: when data of MIME-type ‘hdf/volume’ (indicating data in NCSA Hierarchical Data Format - HDF) was retrieved by the browser, it was passed to VIS as helper application.

The emergence of Java allowed a different approach in which software as well as visualization data was transferred over the Web. An early example of a Java applet for scientific visualization was the NPAC Visible Human Viewer, first developed in 1995 but still widely used today [7].

Wood et al [15] took a different approach again. Rather than transfer visualization data or software across the Web, they proposed a system in which the visualization was executed on a remote server, and the resulting graphics returned over the Web for viewing in a browser. An air quality visualization service was developed in which a user enters details of data and visualization technique on a form; the form is processed by CGI script which invokes the IRIS Explorer visualization system on a server; and returns VRML to the browser - see Figure 1.

In Figure 2 we try to reflect the distinction between these three approaches. They can be seen as separate threads, differing in what is sent over the Web to the browser - either visualization data, or software, or graphics - and differing too in the additional software required at the client side.

We now examine how each thread has developed, and indeed how the threads have intertwined. Examples of the different approaches can be found at our website [14].

Transferring Visualization Data

This thread makes minimal use of the Web, relying on it simply as a data communication medium and exploiting the MIME-typing concept to divert data to a particular application. Users continue to use their own software, on their own platform. The idea has been explored by Hibbard in the Vis5D system [11]. Meteorological data can be tagged with MIME-type ‘application/vis5d’, and a browser configured to pass data of that MIME-type to Vis5D. The user then selects the options in Vis5D to create whatever visualization is required.

This approach can be extended to any turnkey system with fixed functionality. It does not generalise so well, however, to the popular Modular Visualization Environments (or MVEs) such as IRIS Explorer and AVS. These are programmable systems, in which the user connects modules together in a dataflow pipeline (using either a visual editor or a scripting language). Here one wants to be able to send not only the data, but also the specification of the pipeline to process it. We have prototyped this idea using IRIS Explorer as MVE [4, 14]. The server hosts an instruction file specifying the data to be visualized, and the initial pipeline to be created. A typical instruction file will have commands to fetch a dataset, fetch an IRIS Explorer ‘map’ (i.e. a pipeline of modules) and launch IRIS Explorer. On receiving this file, the browser invokes an application that interprets the file and executes the commands. The command set can also include a series of instructions to control the subsequent interactive session, changing parameter values on modules and so on. This opens up some interesting opportunities: for example, it could be used in a consultancy context to deliver a visualization solution over the Web to a customer. They would download the instruction file and a presentation would be automatically executed. This is similar to the ‘pilot’ and ‘passenger’ model used in FAST expeditions [3].

Notice that this approach of sending data, and the pipeline to process it, is an intertwining of the data and software threads in Figure 2. The software framework and the modular components of the MVE are assumed resident on the client, but the programmable connections arrive over the Web. We see a similar intertwining in the next section.

Transferring Software

This thread is the Java applet approach whose origins we can trace to early examples such as the NPAC Visible Human Viewer [7]. These examples were designed for cases where the data was closely associated with the software, and indeed located on the same server. (So in these early cases, the ‘data’ and ‘software’ threads were really intertwined.) However for many applications this is inappropriate - the data usually is associated with the user rather than the software provider. One of the first instances of an applet to process user data was VizWiz [6] - an applet for isosurfacing. This circumvented a Java security requirement that insists an applet can only read files on the host from which it is downloaded. VizWiz uses the browser file upload facility to upload data from client to the server as a temporary arrangement, for the applet to then retrieve it back again. This data transfer can be an issue for large datasets.

We have experimented with similar applet-based approaches, extending the concept to allow visualization of data from any location on the Web. This is achieved by having a ‘data server’ running on the applet host, which fetches the data from a specified URL [14]. There remains the problem of data transfer to and from the server. Today applets can be ‘signed,’ giving them authority to read the local filestore. This will avoid the data transfer overhead, and essentially separates the ‘software’ and ‘data’ threads to give a pure ‘software’ thread. Another important development is Java3D, which will allow visualization applets to include scene graph based rendering. Indeed this can be taken further with the emergence of VisAD [12], a visualization system developed as a set of Java classes - so applets can be built from the visualization system directly.

There is an interesting way to combine the Java applet idea with the concept of MVEs. We have experimented with the idea of embedding Java classes within IRIS Explorer modules, allowing the Java code to be fetched from a central repository at run time [14]. This general area of using Java in a component-based approach is developing fast with growing use of Java beans.

Transferring Graphics

This thread originated with the server-based system of Wood et al [15]. Again it is a very general approach, and several similar systems have been proposed. For example, the authors of the VTK visualization toolkit [8] explain how to build a VTK server-based visualization service.

The form-based interface of these systems is simple but inflexible. Trapp and Pagendarm [9] in their Vis-a-Web system replace the form interface with a more flexible Java applet interface. This intertwines the ‘software’ and ‘graphics’ threads - the user interface being downloaded as software, the visualization as graphics.

A logical extension of this approach in the case of MVEs is to split their operation into two parts: a Java-based front-end that can execute on the client, generated automatically from the pipeline of an MVE application which runs on a server. Again we have experimented with this idea in the context of IRIS Explorer [14], as has Treinish [10] for IBM Data Explorer.

What Next?

There remain many ideas to explore - either as extensions of the threads, or an intertwining of threads. We look at these under the headings:

• Interaction - to involve the user more.
• Collaboration - to support group working.
• Monitoring - to visualize data on-line.

Interaction

In the server-based approach, where the visualization is generated remotely and only graphics returned, interaction is often limited. However combining the ‘graphics’ and ‘software’ threads, we can program interaction into the visualization. This approach has been tried in Web-based virtual reality applications in the medical area [5, 2] where simple surgical simulations combine VRML representations of anatomy and instruments, with interaction code in Java - linked through the Java EAI for VRML. The same approach could be used in data visualization to allow more complex interaction than afforded by VRML on its own.

VRML itself is evolving. Its successor, X3D [13], uses XML and defines a core set of nodes - plus an ability to include new nodes. Thus it may be possible to include visualization controls alongside the geometry capabilities.

A different approach for the ‘graphics’ thread is to do all the rendering on the server side and simply transfer images to the client. One could envisage a simple Java applet on the client being used to control the camera. This can take advantage of high quality rendering software provided on a server to give better image quality than a VRML viewer. This close coupling of visualization and rendering may also improve the ease of interacting with the underlying data - in the VRML model where visualization and rendering are distributed, one has to send data as well as graphics in order to support interrogation of data values. Here data, visualization and rendered image would be co-located.

Collaboration

Collaborative Web-based visualization allows groups of users to build on each other’s work. We have experimented with this for the server-based approach [16]. The IRIS Explorer air quality visualization system was given a measure of persistence, by allowing a user to have their parameter selection on the form stored in a database on the server. This selection can then be made visible to subsequent visitors to the site. In fact a sequence of ‘snapshots’ can be recorded, forming a tree of exploration. This is illustrated in Figure 3 where we see the tree built up over a period of collaborative study.

Monitoring

There are a growing number of applications where the data being visualized is constantly changing - whether it is stock market prices or patient heartbeat (which may indeed be correlated!). This will also occur when we link visualization on line to a simulation, for example in computational fluid dynamics. One could envisage here a Java applet approach, in which the applet has a direct socket connection to a constantly updated data source.

Final Thoughts - Intertwining and Disentangling the Threads

The Web has had a profound effect on almost all aspects of computing, and visualization is no exception. The three original threads of ‘data,’ ‘software’ and ‘graphics’ continue to develop, to intertwine and to disentangle. The combination of Java applets acting as front-ends to a visualization system on a server intertwines the ‘software’ and ‘graphics’ thread; if the applet also retrieves data for some simple analysis at the client side, then all three threads become intertwined. We are still learning different ways to harness the potential of Web-based visualization, and the examples mentioned here are only a selection of what has been tried to date.

We are building a live repository of examples of Web-based visualization on our website - http://www.scs.leeds.ac.uk/ vis/webvis. Please let us know if you have an example we can add.

Acknowledgments

Many students have helped us over the years in our understanding of Web-based visualization. Special thanks are due to Abraham Kee, Peter Stanton, Edward Teong, Alan Yeo, Alex Coyle, Nuha El-Khalili and Ying Li, who have prototyped some of the ideas described here and whose work is summarised at our website. Much of the work has been carried out within the University of Leeds IRIS Explorer Centre of Excellence.

References

1.  Ang, C.S., D.C. Martin and M.D. Doyle. "Integrated control of distributed volume visualization through the World Wide Web" in R.D. Bergeron and A.E. Kaufman (editors), Proceedings of IEEE Visualization94, pp.13-20, 1994.
2.  El-Khalili, N.H. and K.W. Brodlie. "Architectural design issues for web-based virtual reality training systems" in P. Fishwick, D.R.C. Hill and R. Smith (editors), Proceedings of 1998 International Conference on Web-based Modelling and Simulation, pp. 153-158, Society for Computer Simulation International, 1998.
3.  FAST. See the FAST project website, 1999.
4.  IRIS Explorer. See NAG Ltd website, and University of Leeds IRIS Explorer Centre of Excellence website.
5.  John, N., N. Phillips, R. Vawda and J. Perrin. "A VRML simulator for ventricular catheterisation" in Proceedings of Eurographics UK Conference, Cambridge, UK, 1998.
6.  Michaels, C. and M. Bailey. "VizWiz: A Java Applet for Interactive 3D Scientific Data Visualization on the Web," Proceedings of IEEE Visualization97, pp. 261-268, ACM Press, 1997. See also VizWiz home page, website.
7.  NPAC. The NPAC Visible Human Viewer website, 1999.
8.  Shroeder, W., K. Martin and W.E. Lorensen. The Visualization Toolkit: An Object-Oriented Approach to 3D Graphics, Prentice-Hall, 1996. (See also vtk website)
9.  Trapp, J.C. and H.G. Pagendarm. "A Prototype for a WWW-based Visualization Service" presented at 8th Eurographics Workshop in Visualization in Scientific Computing, Boulogne sur Mer, 1997.
10.  Treinish, L. "Enablement of a Fluid Client-Server Visualization Environment," Workshop on PC-Based Visualization and Computer Graphics, IEEE Visualization97, 1997. (See also website)
11.  Vis5D. Vis5D website. See: website, 1999.
12.  VisAD. VisAD website. See: website, 1999.
13.  Web3D. Web3D Consortium website. See: website, 1999.
14.  Webvis. Web-based Visualization. See: website, 1999.
15.  Wood, J.D., K.W. Brodlie and H. Wright. "Visualization over the WWW and its application to environmental data," Proceedings of IEEE Visualization96, pp. 81-92, 1996.
16.  Wood, J.D. Collaborative Visualization, Ph.D. thesis, University of Leeds, 1998.

Ken Brodlie, Stuart Lovegrove and Jason Wood
School of Computer Studies
University of Leeds
Leeds, UK, LS2 9JT


Bill Hibbard
Space Science and Engineering Center
1225 W. Dayton Street
Madison, WI 53706

Tel: +1-608-253-4427
Fax: +1-608-263-6738
Website


The copyright of articles and images printed remains with the author unless otherwise indicated.