New Approaches to Scientific Visualization over the Internet

One of the most important developments in communication technology during the past decade has been the evolution of the World Wide Web (WWW) based on the Internet. The combination of easy-to-use web browsers on multiple platforms and distributed servers provides a powerful facility to distribute information to a large number of users. We see the Internet as an excellent foundation to connect new developments in the field of scientific visualization and a large community of users, almost regardless of their hardware and software ressources. Scientific visualization ("creating images out of numbers") is typically used to investigate or illustrate huge data sets by researchers working in many disciplines. An important topic within the field of scientific visualization is the visualization of dynamical systems. The behavior of many real world phenomena like the development of stock market rates, weather patterns, wave propagation, econometric systems, chemical reactions, or biological processes, is often modelled as a dynamical system. The understanding of such models provides useful insights into the behavior of the real world system or related problems. Scientific visualization helps to investigate and analyze complex dynamical systems, which usually is rather difficult due to, e.g., multiple dimensions, non-linearity, or chaotic behavior. By applying advanced visualization techniques, the properties of the system can be depicted in a more intuitive and easily understandable way. A large number of visualization techniques has been developed for this purpose, ranging from icons and glyphs depicting the behavior at certain states of the system to methods representing large scale evolutions over time, like streamlines or streamsurfaces. The development of such visualization methods is usually carried out by scientists using specialized graphics hardware (for example Silicon Graphics workstations) and dedicated visualization software like AVS or IRIS Explorer. A majority of infrequent users has either no access to the necessary hardware and software, or no time to become familiar with the handling of a complex visualization package. Thus the transfer of innovations from scientific developers to a broad field of users is usually slow and inefficient. We propose to exploit the portable and transparent computing opportunities of the Internet and the ease-of-use of the WWW to bring the benefits of advanced visualization techniques to users working with dynamical systems without access to expensive hardware and software resources. To be broadly accepted by the user community, a good solution must include several important features: Portability: The system should be portable and run on various platforms ranging from UNIX workstations to desktop PCs. Java and VRML provide the power to realize these features. Ease-of-use: The system offers a set of predefined visualization configurations covering common applications and presents the same intuitive and easy-to-leam interface on each platform. Flexibility: The system should be flexible, allowing the user to visualize his or her data in a customized way with a minimum amount of work. Interactivity: The investigation of the user data should be interactive, almost regardless of the user`s desktop computer. Computation tasks may be distributed between client and server. Both the users and the developers benefit from this approach. The users apply improved visualization techniques to investigate their problems without big expenses using a simple and easy-to-learn interface. The scientists developing new visualization techniques receive rapid feedback about the usefulness of their methods from a broad field of applications.

"Visualization over the Internet does the work where the processing power is and transmits the visual results to where the user needs it" One of the most important developments in communication technology during the past decade has been the evolution of the WorldWideWeb (WWW) based on the Internet. The combination of easy-to-use web browsers on multiple platforms and distributed servers provides a powerful facility to distribute information to a large number of users. The aim of the project (BandViz for short) is to deliver quality interactive visualization via the Internet bridging the gap between different research groups. Visualization applications are characterized by huge data sets and high computational demands. Distributed visualization over the internet uses a network of interconnected computers transparently as one extremely powerful processing entity. The various steps in the visualization pipeline (data acquisition, filtering, mapping, image generation) are thereby done at those processing nodes where it is most appropriate. Issues are large data handling and transmission. Within the project the following topics where addressed: Efficient data representation: Compression of large data sets, e.g., volume data, is a crucial step for visualization over the Internet. A progressive and hierarchical representation allows progressive transmission and enables sustained frame-rates. This is especially interesting for extremely large medical volume data sets. Novel visualization techniques: internet-based visualization is still characterized by output devices with greatly varying performance capabilities. Therefore rendering styles are necessary which allow even on low-end hardware a fast representation of the most important data aspects. We developed two-level volume rendering which is a new approach to integrate different rendering styles. Also non-photorealistic styles like outline and silhouette rendering turned out to be promising methods for volume data. Econometric models (describing interactions between competing companies) required the development of novel visualizations for illustrating complex structures like attractors and basin-boundaries. Salient feature extraction: With the rapidly increasing size of data sets it is crucial to concentrate on the most important part of the data. By analysing higher-order derivatives we are able to extract salient features, which typically require much less storage and transmission capacity. In the area of medical visualization large data sets are reduced through intelligent segmentation algorithms which e.g., eliminate bony structures and extract vessel trees. In summary internet-based visualization has to cope with large data and varying hardware capabilities. In the BandViz project we concentrated our efforts on data compression, extraction of the most important data features, and scalable rendering styles (two-level volume rendering, non-photorealistic rendering). The concepts were evaluated through collaborations with medical doctors and economics scientists.