=Paper=
{{Paper
|id=None
|storemode=property
|title=Declarative 3D Approaches for Distributed Web-based Scientific Visualization Services
|pdfUrl=https://ceur-ws.org/Vol-869/Dec3D2012_07.pdf
|volume=Vol-869
|dblpUrl=https://dblp.org/rec/conf/www/JungBDW12
}}
==Declarative 3D Approaches for Distributed Web-based Scientific Visualization Services==
https://ceur-ws.org/Vol-869/Dec3D2012_07.pdf
Declarative 3D Approaches for Distributed Web-based
Scientific Visualization Services
Yvonne Jung Johannes Behr
Fraunhofer IGD Fraunhofer IGD
Darmstadt, Germany Darmstadt, Germany
yjung@igd.fhg.de jbehr@igd.fhg.de
Timm Drevensek Sebastian Wagner
Fraunhofer IGD Fraunhofer IGD
Darmstadt, Germany Darmstadt, Germany
tidreven@igd.fhg.de sewagner@igd.fhg.de
ABSTRACT Android, RIM, WindowsPhone), different desktop systems,
Recent developments in the area of efficient web-service ar- cloud bases etc., which typically make cross-platform devel-
chitectures and the requirement to provide applications not opment complicated and time consuming. In this regard,
just for a small expert group lead to new approaches in web technologies incl. web-service architectures are rather
the field of web-based (scientific) visualization. The just common and can overcome these issues by interconnecting
emerging support for GPU-supported and therefore high- all these diverse approaches. Moreover, web browsers are
performance 2D and 3D graphics in modern web-client im- converting more and more towards full runtime environ-
plementations and standards provide new application envi- ments for whole applications and with WebGL or Stage3D,
ronments, which are especially interesting for the demands 3D-capable browsers are now broadly available, too.
of scientific visualization solutions. Thus, in this paper we
present a web application deployment architecture that aims However, this does not solve the often intricate processes
at supporting decision making processes more efficiently. We to develop applications, where one can distinguish between
also show that current approaches in the field of declarative two main types. For one thing, we have interactive processes
3D techniques are useful for client-side rendering as well as with creative and individual components such as games with
for a large number of processing and visualization aspects. their manual or semi-automatic tool chains (e.g., DCC tools
like 3ds Max and game editors like Unity3D) that focus on
isolated applications. For another, there are automatable
Categories and Subject Descriptors processes that typically are complex and distributed, which
H.3.5 [Information Systems]: Online Information Ser-
can be realized by means of a web-service architecture. In
vices—Web-based services; I.3.6 [Methodology and Tech-
the context of this paper, we focus onto the latter approach
niques]: Standards—Languages; I.3.7 [Computer Graph-
by presenting a fully automated Web Service Portal, which
ics]: Three-Dimensional Graphics and Realism—Virtual Re-
thereby allows generating e.g. 2D/3D mash-ups for design
ality
review, city planning or similar decision making tasks.
Keywords Generally spoken, distributed data-centered applications are
X3D, HTML5, WebGL, DOM, Web Integration, Transcoder, one of the common implementation concepts for scientific vi-
Web Services, Service-oriented Architectures, X3DOM sualization solutions today to process huge data sets. There-
fore, utilizing web-service architectures or even cloud-based
1. INTRODUCTION solutions are just the next logical step. Recent develop-
With the advent of 3D Internet, there is a strong trend to- ments in the area of high-performance web-service archi-
wards 3D data and documents (like 3D scanners and print- tectures and the requirement to provide applications not
ers, geo-referenced data, and Augmented Reality) as well as just for a small expert group lead to new approaches in
a convergence of application platforms (e.g. W3C WebApps, the field of web-based visualization. The emerging support
HTML5, etc.) towards web-based contents. However, we for hardware-accelerated and therefore high-performance 2D
still have various operating systems with very different soft- and 3D graphics in modern web clients and standards pro-
and hardware requirements such as smart phones (e.g. iOS, vide a new application environment, which is especially in-
teresting for the demands of scientific visualization solutions.
Therefore, some visualization packages (like ParaView with
ParaViewWeb from Kitware1 ) already use this opportunity
to move the established application model to a web and
cloud-based solution. Deploying current application models
Copyright c 2012 for the individual papers by the papers’ au- to a new environment is just the first step. However, this
thors. Copying permitted only for private and academic purposes. field is not yet explored to utilize its full potential.
This volume is published and copyrighted by its editors.
1
Dec3D2012 workshop at WWW2012, Lyon, France http://paraview.org/Wiki/ParaViewWeb
challenging application domains: medical imaging and radar
meteorology. Performance, scalability, accuracy, and secu-
rity are some of the many challenges that must be solved
Figure 1: Coarse sketch of the visualization pipeline. for 3D web applications, esp. since WebGL is still based on
the old Shader Model 2.0 (like on the NVidia GeForce 5900
series introduced in 2003).
Thus, the web-application deployment component presented
in this paper helps supporting decision making processes
Another alternative method for representing 3D graphics is
more efficiently, and the benefits include the following four
using point clouds, which is a set of points in 3D space with
major aspects:
attributes. This is especially of interest, since 3D scanning
– leveraging new GPU-accelerated 2D and 3D client APIs;
devices such as LiDAR equipment and sonar scanners de-
– achieving interactivity while combining traditional client-
liver their data as point clouds. One recent framework to
and server-side rendering techniques;
simplify the streaming and rendering of point clouds based
– providing new user experiences while considering scalabil-
on JavaScript and WebGL [15] is “XB PointStream” [20].
ity and security in open web-based environments;
– automating and optimizing data preparation processes for
In [22], the authors propose a WebGL-based framework for
web-specific environments through web server infrastructures.
representing reflectance information via Bidirectional Tex-
ture Functions (BTF), which allows for the progressive trans-
2. RELATED WORK mission and interactive rendering of digitized artifacts con-
For the design of the proposed web-service architecture sev- sisting of 3D geometry and reflectance information. This
eral domains and technologies are relevant. Corresponding is achieved by employing a novel progressive streaming ap-
work is thus briefly described within this section. proach for the huge BTF data set that allows the smooth
interactive inspection of a steadily improving model during
2.1 Scientific Visualization download. Analogously, the X3DOM framework [3, 4] cur-
To provide important features for scientific visualization (cf. rently follows a similar approach for lightweight geometry
e.g. [6] and [14]) in a specialized web application, the appli- compression and transmission via so-called image geometries
cation must be able to handle visualization-specific represen- that likewise utilizes image compression techniques.
tations that consist of registered and merged point, surface,
and even volume data as well as the corresponding meta in- 2.3 Declarative (X)3D in HTML5
formation. But having the raw data to be visualized readily The open ISO standard X3D [8] provides interactive 3D
available is only the very first step, as is shown in Figure 1 graphics for the web and is the only standardized 3D de-
with a rather simplified visualization pipeline. Usually, after ployment format. It differs from other 3D formats like the
data acquisition the information is re-sampled onto a struc- interchange format Collada [1] in that it also includes the
tured regular grid before filtering the data accordingly (a scene’s runtime behavior. The X3D standard already pro-
discussion of common data representations including multi- vides point and surface primitives. Typically, the visual el-
resolution and adaptive resolution representations for large ements are described by their boundary representation, e.g.
data sets can be found in [14]). Then, the scalar, vector, or via an “IndexedTriangleSet” node. Volume rendering will be
tensor values of the data set are mapped to visual represen- part of the next spec revision. Therefore, the Web3D med-
tations that can be rendered [6]. Well-known examples here ical working group has presented a sample implementation
are volume rendering or flow visualizations in CFD [17]. [11] and an ISO/ IEC PDAM extension for a volume render-
ing component in X3D, which also has answers to DICOM
For multi-dimensional, multi-variate/ multi-modal, and prob- requirements for n-D presentation states [19].
ably even time-varying data, visualization is much more in-
tricate, since first the registration between different data In addition, since most geo-referenced data is provided in
sets including the thereby introduced uncertainties has to be a geodetic or projective spatial reference frame, the X3D
handled as well as data fusion aspects and the calculation Geospatial component includes conventions that are defined
of derived quantities (e.g. the strength of correlation be- by the Spatial Reference Model and thereby provides sup-
tween two different variables) to improve the visualization port for geographic and geospatial applications [8, 16]. With
[6]. In this regard, not only the choice of an appropriate surface, volume, and geo-spatial components X3D already
visualization technique but also the possibility to interac- provides a solid foundation for scientific visualization tasks
tively explore the data set (e.g. by changing the value range and thus enables an automated connection of existing data
or transfer function) is of great importance for grasping and with e.g. atmospheric, oceanographic, or geological data to
analyzing the information and therefore requires real-time be visualized in the Web.
capable mapping and rendering methods, which nowadays
is often achieved using GPU-based methods [17]. However, all existing and proposed X3D components only
extend X3D for various application and visualization sce-
2.2 3D Graphics in the Web narios but do not address one of the essential issues of X3D
Direct volume rendering in general is an alternative form right now, namely that X3D is still bound to a plugin-based
of visual data representation compared to the traditional integration model that has major usability and performance
polygonal form. In [7] the implementation of a direct vol- issues especially for large data sets [3, 4]. All plugin-based
ume rendering system for the Web is presented. In this systems have two major drawbacks. First, they are only plu-
publication the authors discuss their WebGL-based volume gins and not installed by default on most systems. There-
rendering approach using ray-casting in two different but fore, the user has to deal with plugin installation, security,
and browser or OS incompatibility issues, not to mention MPEG video streaming. Smart environments consisting of
the lack of accessibility. Second, the presented systems de- several interconnected devices, which can change dynami-
fine an application and event model inside the plugin, which cally whenever mobile devices enter or leave the environ-
is decoupled from the DOM content. Developers, who try to ment, are specifically addressed in [24]. To utilize such en-
develop integrated web applications that use both, the DOM vironments efficiently for information visualization, the au-
and the plugin-model, have to deal with small plugin-specific thors also propose a service-oriented architecture. Although
interfaces and synchronization problems. their work still focuses on the specificities of certain display
devices, it nevertheless already also addresses issues such as
WebGL (a JavaScript binding for OpenGL ES 2.0 in the caching strategies and compression methods.
Web Browser [15]), Adobe’s Flash 11 with Stage3D [23],
and Microsoft’s Silverlight 5 [18] now all provide access to 3. SYSTEM ARCHITECTURE
the native GPU layer without any further plugin installa- Within this section, we focus onto automatable processes for
tion process, but the issue of the missing DOM integration web application development by designing and implement-
still exists. Hence, what is still missing is a better integra- ing a fully automated Web Service Portal, which provides
tion with open architectures, which integrate well with ex- services that automatically combines application templates
isting web technologies like CSS(3), HTML(5), JavaScript, and raw data into interactive 3D visualizations for the web.
DOM scripting and Ajax. In this regard, with the X3DOM
project recently a DOM-based integration model for declar-
ative (X)3D in HTML5 was proposed [3, 4], that allows a
3.1 Use Cases and Design Considerations
The goal here is to provide web services that automatically
seamless integration of 3D contents into the HTML docu-
convert raw 3D data (e.g. geo-spatial data, simulation re-
ment model by utilizing standard web APIs for integrating
sults, and architectural models) into interactive 3D visual-
content and interactions.
izations for the Web that can be delivered as a cloud service.
For example, value ranges that can be interactively explored
2.4 Service-oriented Architectures like a temperature or density distribution are automatically
Finally, we have to shortly review service-oriented architec- tagged, like e.g. it is often done in modern IDEs such as
tures (SOA). An SOA is a software architecture paradigm the CG FX Composer framework, to provide all data that
for structuring and using distributed functionalities that are is necessary for an appropriate user interface, like a slider
managed by various owners. Comparable to business trans- with a corresponding value range as shown in Figure 4.
actions an SOA composes several low-level services to more
complex services with a higher level of abstraction [9]. Here, This enables the user (in our case typically a decision maker)
web-services help supporting the collaboration between dif- to interactively explore the presented 3D data, which is
ferent applications running on different platforms by uti- achieved with the help of two approaches. On the one hand,
lizing REST, JSON, and XML-based standards like UDDI, certain data values and ranges are reflected in such a way,
WSDL, and SOAP as web-service protocols. In this regard, that the application service can directly derive appropriate
one prominent application of web-services are the standard- UI elements (e.g. a slider from x meters to y meters that
ized geo-services (such as WCS [2]) provided by the Open allows visualizing different water-levels in a disaster man-
GeoSpatial Consortium (OGC) to make spatial geo-data ac- agement scenario). On the other hand, the user must be
cessible in a structured way. able to directly interact with the visualization, e.g. by se-
lecting a certain region for which he wants to obtain more
Hence, some 3D rendering systems already make prototyp- information (which e.g. in a GIS scenario might be “show
ical use of REST and JSON for scene-graph manipulation buildings and trees in marked area”) or by clicking onto a
and distributed rendering [21] within a service-oriented plat- certain POI, which in turn delivers other information from
form. To demonstrate the full potential of their web-service the processing backend.
interface, the authors additionally included a second visual-
ization backend that provides global illumination algorithms Therefore, the visualization needs to be scalable in such
that are evaluated on a server farm whereas the web frontend a way that – depending on the desired application type –
only shows the final image and the UI elements. Likewise, mobile and desktop machines with rather different comput-
in [5] a full workflow pipeline from data acquisition up to ing and 3D capabilities are supported alike. For example
interactive web-based visualization in the Cultural Heritage this can be achieved via a hybrid approach that provides
domain was proposed. Remote or server-based rendering in both, server-side streaming for low-end mobile clients (us-
general is a very active research discipline for the last 20 ing InstantPlayer as render server [10]) and direct web-based
years. Visualization servers or services are build to over- 3D rendering for desktop machines via X3DOM. Moreover,
come the performance issues with client solutions or provide cross-platform and cross-browser issues need to be addressed.
specific systems to protect the 3D dataset while not or only Finally, especially when also CAD and PDM or highly clas-
partially distributed to the client. Koller et al. [12] provide sified data is used, security aspects are of high importance.
an interesting proposal to the protection system with a re- Here again, streaming technologies help with information
mote rendering service that not just only transfers images hiding in that only image streams but no 3D models are
to the client but also includes a number of active defense transferred over the network to the client for display.
method to guard against 3D reconstruction attacks.
As mentioned, the proposed software architecture is based
A streaming-based remote visualization for mobile devices on web services. As described in section 2.4, these are in-
was presented in [13], where complex 3D models are ren- tended for being used in broader software ecosystems, which
dered on a cluster of PCs and transferred to the client via automatically exchange data. Usually, the service interfaces
Figure 2: Instant3D Hub – scalable, secure, and efficient SOA for 3D visualizations on the Web.
are described by XML documents such as SOAP. Obviously, template provides at least the HTML page, the necessary
utilizing a unified standard for shared services and commu- meta-information of the respective application, as well as de-
nication is essential here. Furthermore, declarative XML- scriptions of all required data containers, e.g. in the declar-
based languages such as X3D are suited well for SOAs in ative X3D format for 3D contents.
that 3D content, such as CAD/ CAM data or a given WCS
response in a GIS application, can be directly transformed, After that, the Hub invokes the appropriate transcoder ser-
e.g. via an XSLT or similar transform, from one representa- vice for providing the concrete data (like a 3D world with
tion to another (such as an X3D world that can be rendered a certain camera pose). Therefore, the transcoder service
in real-time in the web browser using the X3DOM frame- translates and, where required, caches the 3D data from
work). As shown in Figure 2, this transformation is done already existing descriptions by converting them from var-
with the help of a transcoder service, which also needs to ious resources and formats, which are not yet suitable for
consider the respective client capabilities (e.g. mobile device real-time web presentation, to a declarative deployment for-
vs. desktop system) to provide an appropriate visualization. mat for final presentation. Here we use X3D in combination
with HTML5, since declarative formats nicely fit to SOA
For being able to present even large-scale big data, suitable pipelines. The result is then provided under a new URI.
compression schemes are required for both, the client (esp. This basic architecture tries to fulfill three basic require-
when considering mobile devices) and the server (where esp. ments that are outlined next.
latency is an issue). In our implementation this is handled by
utilizing image-based compression and delivery algorithms
that can encode geometric as well as material data. In ad- Scalability The central HUB recognizes the browser crite-
dition, appropriate server-side caching strategies allow for ria (e.g. mobile or desktop) and can use those to ask
faster delivery of 3D contents. for a device- and/ or context-specific application UI.
The application uses a scene-graph manipulation layer,
which maps the actual changes to browser local (e.g.
3.2 Web-Service Architecture DOM content) or remote scene representations.
Figure 2 coarsely shows the concept of the proposed service-
oriented architecture. The core component is the “Instant3D Efficiency The overall structure is build to produce very
Hub” that acts as application provider, whereas the appli- interactive and pleasant user experiences. All main UI
cation identification is resolved via a URI. Thereto, the ap- pages should get delivered immediately and all further
plication service takes a request from the so-called Hub and data sets should be provided with a flexible caching
extracts an application template, which specifies the data infrastructure on request. The whole service architec-
and application characteristics. For data preparation, the ture is build asynchronous and does not wait at any
point for any data conversion process at all. The client head includes a standard Retry-After element that de-
and server side rendering processes are both built to fines when the resources is expected to be available.
be able to stream efficient binary updates to the server
for highly responsive environments.
The transcoder also includes an efficient caching and updat-
Flexibility The architecture is not bound to a specific ap- ing mechanism, which allows to convert the data only once
plication model and very open through the web-service for similar requests, but also updates the converted data as
infrastructure. It provides an environment useful to soon as new data sets are available.
a large number of applications by harmonizing three
basic steps: application-UI retrieval, continuous data- 3.5 Rendering
preparation and a hybrid visualization approach, which The converted data is fetched to a real-time rendering sys-
can map to client and server side rendering. tem, which consists of a scene-graph whose elements can be
identified and modified during runtime. This scene-graph
3.3 Application and UI retrieval is manipulated through a client-side abstraction layer that
The goal of the framework is to deliver context- and browser- allows mapping those changes to local or remote represen-
specific web applications. The application service provides tations of the graph.
templates for those applications, which are retrieved on re-
quest. Those templates consist of the following parts: 3.5.1 Client-side Rendering
Depending on data security aspects and the respective web
browser capabilities, the Hub generates a concrete visual-
HTML documents HTML documents at first can be seen ization application from a given pre-prepared scenario tem-
as the building ground for all application-specific struc- plate. For one thing, this can be realized via client-side
tures. Those documents are provided by the applica- rendering by utilizing X3DOM, which either uses JavaScript
tion service for a specific browser and therefore device with WebGL or Flash 11 with Stage 3D for real-time render-
(e.g. tablet or desktop). ing. Open research issues here are suitable caching strategies
Data references The application includes usually a list of for fast content delivery as well as scalable methods for bi-
external resources, which should be delivered to the nary compression of high-end material data (compare [22])
browsers while running. This could be resources like and of big geometric data (i.e. vertex attributes).
XML documents or images, which do not need any
further processing but also link to data sets that need The latter can be achieved by utilizing our image geometry
to be converted before viewed (e.g. a specific 3D data approach. This method not only allows to asynchronously
set) or transformed for a specific device class. deliver and compress vertex attribute data but it also allows
to nicely separate the structure of the 3D content from raw
Metadata The metadata describes criteria for the conver- vertex data that usually only bloats the HTML document.
sion and transcoding process. This can e.g. control The advantages of client-side rendering are a rather simple
whether optimizations on the graph are allowed or not server infrastructure and highly interactive apps since ev-
and how identification of objects should be performed. erything is rendered on the client. Disadvantages are that
the data-load can easily overburden the client, that lots of
3d data needs to be transferred, and security or IPR issues.
The hub requests this information from the application ser-
vice, requests then data sets from the transcoder service
with the data references and metadata settings and delivers 3.5.2 Server-side Rendering
the final HTML document with the not yet existing external The second option is server-side rendering. In this case, for
references. instance an X3D-based runtime environment [10] is utilized
for rendering and the rendered image frames are transferred
(e.g. as an MJPEG stream) to the client. Since interactions
3.4 Transcoder are handled via WebSockets and XMLHttpRequest (XHR),
The transcoder service converts the data of a given resource the latency is much higher than for client-side rendering.
to a new single resourc or set of resources while considering Therefore, we are currently exploring suitable interaction
the guidelines provided by the application metadata. The methods and message protocols.
resulting URI’s are provided immediately as result of the
transcoding request even so the data may not be available Another disadvantage here is the fact that a complex server
yet. The basic HTTP return code is used to communicate infrastructure is required. However, since no 3D data but
the availability of a specific resource while requesting. only images are transferred over the network, there are no
IPR issues concerning the 3D data, and data security is in-
herently given as a nice side effect. Also, the visualization
200 OK: the request has succeeded and the entity corre-
application can be executed on arbitrary clients, which is
sponding to the requested resource is sent in the re-
another advantage of utilizing server-side rendering.
sponse.
404 Not Found: the conversion job could not be completed 4. RESULTS AND DISCUSSION
successfully and the sources will not be available with- 4.1 Application Examples
out it. In this section a few use cases are exemplarily described to
503 Service Unavailable: The service is not yet available demonstrate the potential of the proposed service architec-
since the conversion process is still running. The HTML ture. In the area of cultural heritage for instance, working
highly specialized and rather sophisticated, they still utilize
proprietary formats and methods that are neither compati-
ble in their concepts of operation nor in their supported data
formats. On the one hand this prevents a harmonization of
data from different sources and thereby hinders its distribu-
tion and utilization. On the other this also leads to parallel
developments of incompatible and isolated technologies.
In this context, the open ISO standard X3D [8] is the only
standardized 3D deployment format that also defines the
runtime behavior. Hence, there are ongoing efforts [19] to
develop an open interoperable standard for the representa-
Figure 3: Presentation of laser scanned CH artifacts tion of volumetric data based on input from a wide variety
using our proposed web-service architecture. of modalities. The proposed X3D volume rendering compo-
nent [8] therefore aims at the exchange and interactive ex-
ploration of volumetric data and at industrial applications
that use X3D as interchange format, but can link to propri-
etary databases and hardware, too.
A service for open, flexible, and scalable access and pro-
cessing of Earth data is the OGC Web Coverage Service
(WCS) 2.0 Standard, which now allows providing a com-
prehensive portion of Earth science data categories through
one coherent and implementation-independent interface [2].
The coverage model of WCS 2.0 transcends pure raster data
Figure 4: Picking of generic vertex and/or fragment and includes almost all relevant categories, such as irregular
properties (left: material thickness of a structural and curvilinear grids, general meshes, trajectories, surfaces,
element, right: electromagnetic field strength). solids, and point clouds.
In this regard, X3D already incorporates basic means for
with 3D scanner data for preservation is getting more and point rendering, as well as a geospatial component. Since
more common. Figure 3 shows an example of a web appli- most geo-referenced data are provided in a geodetic or pro-
cation for the visualization of scanned historical 3D objects, jective spatial reference frame, X3D therefore provides sup-
which was generated using the proposed transcoder architec- port for a number of nodes that can use spatial reference
ture by automatically processing and converting the scanned frames for modeling purposes. However, there are still sev-
data to create an interactive viewer. eral drawbacks like the lack of well-defined terrain rendering
etc., which for instance were addressed in [16]. In addition,
Figure 4 shows a 3D CAE prototype for the visualization with the X3D Earth working group there is a strong collab-
of simulation data in the web browser using X3DOM [3, 4], oration of the Web3D Consortium with the OGC, because
where two examples from the domain of sheet metal forming as mentioned, X3D provides a solid foundation here and is
(left) and electromagnetic field simulation (right) are taken a good starting point for further standardization efforts to
to allow the exploration of FEM results. These application enable an automated connection of existing data sources for
prototypes have a huge potential for communicating and further visualization.
presenting e.g. simulation results towards decision makers
or consumers, without distributing a whole application while Hence, a pro-active participation in several standardization
providing the user with more information than a static image bodies (ISO, W3C, Khronos) helps preventing parallel devel-
or a boring fact sheet. opments of isolated technologies by harmonizing the general
understanding. Therefore, the “Declarative 3D for the Web
The example to the left in Figure 4 shows the results of a Architecture” W3C Community Group was founded in 2011
sheet metal forming simulation in automotive environments. with the objective to standardize a declarative approach to
The material thickness after the forming process is color- interactive 3D graphics as part of HTML documents.
coded and applied via a look-up texture. By clicking onto a
colored region, the corresponding thickness value is obtained 5. CONCLUSIONS
and marked with a yellow arrow and textual information in Traditional scientific visualization approaches are extremely
the top right of the application. The slider elements below demanding in regards to software and hardware require-
are implemented using the JavaScript library jQuery (http: ments. Thus, today’s visualization solutions are tailored
//jquery.com/) and can be used to interactively modify the for specific data representations and application scenarios
color-coding by setting offset, bias, and threshold. to cope with several soft- and hardware limitations, which
led to extremely specialized software packages that are built
4.2 Standardization Aspects and run by professionals in high-cost scenarios. However, re-
Another point is the lack of open standards in scientific visu- cent developments in the area of web applications and the re-
alization and 3D graphics as well as in mobile and interactive quirement to provide applications not only for a small group
systems. Although current visualization systems are often of experts lead to new approaches for web-based visualiza-
tion. Morever, since data-centered, distributed applications State of the Art Reports, pages 117–134, 2007.
were one of the common concepts to process huge data sets [7] J. Congote, A. Segura, L. Kabongo, A. Moreno,
for visualization, web-service architectures and cloud-based J. Posada, and O. Ruiz. Interactive visualization of
solutions are just the next step. Additionally, the emerging volumetric data with webgl in real-time. In
support for GPU-accelerated high-performance 3D raster- Proceedings Web3D 2011, pages 137–146, New York,
graphics in modern web clients and standards provide an NY, USA, 2011. ACM.
application environment that is especially interesting for the [8] W. Consortium. Extensible 3d (X3D), 2011.
demands of scientific visualization solutions. http://www.web3d.org/x3d/specifications/.
[9] T. Erl. Service-Oriented Architecture: Concepts,
In this regard, the approach we presented in this paper does Technology, and Design. Prentice Hall PTR, 2005.
not only implement an established application model in a [10] FhG. Instant Reality, 2012. www.instantreality.org.
new web- and cloud-based environment, but it also intro-
[11] N. John, M. Aratow, J. Couch, D. Evestedt,
duces a web-service architecture as core of the application
A. Hudson, N. Polys, R. Puk, A. Ray, K. Victor, and
distribution and deployment infrastructure. Instead of sim-
Q. Wang. MedX3D: Standards enabled desktop
ply delivering data to a rigid viewer like in conventional
medical 3d. In MMVR: Medicine Meets Virtual
applications, our proposed architecture delivers a user ex-
Reality, pages 189 – 194, 2008.
perience that is tailored for specific user scenarios and the
actual context. Thereby, the final environment is able to de- [12] D. Koller, M. Turitzin, M. Levoy, M. Tarini,
rive and deliver a dynamic application with regards to user, G. Croccia, P. Cignoni, and R. Scopigno. Protected
system, security, and data requirements. interactive 3d graphics via remote rendering. In ACM
SIGGRAPH 2004 Papers, SIGGRAPH ’04, pages
This allows the system to provide dynamic technical solu- 695–703, New York, NY, USA, 2004. ACM.
tions and to automatically choose a client, server, or even [13] F. Lamberti and A. Sanna. A streaming-based
hybrid visualization method for an unchanged application solution for remote visualization of 3d graphics on
at runtime, while also considering the device capabilities as mobile devices. IEEE Transactions on Visualization
well as IPR and security concerns. What is even more im- and Computer Graphics, 13(2):247–260, 2007.
portant, besides the technical aspects, is the architecture’s [14] D. R. Lipsa, R. S. Laramee, R. D. Bergeron, and T. M.
ability to provide very specific applications for certain user Sparr. Data representation for scientific visualization:
groups, since the system is not built as a rigid application An introduction. In Computer Graphics, chapter 6,
model for a small user group. Thus, the presented service pages 121–136. Nova Science Publishers, 2011.
architecture enables new application scenarios, including sci- [15] C. Marrin. Webgl specification, 2011.
entific visualization for decision support or visual analytics https://www.khronos.org/registry/webgl/specs/1.0/.
applications, as well as e.g. corresponding web applications [16] M. McCann, R. Puk, A. Hudson, R. Melton, and
for public information scenarios. D. Brutzman. Proposed enhancements to the x3d
geospatial component. In Proceedings Web3D 2009,
pages 155–158, New York, NY, USA, 2009. ACM.
6. ACKNOWLEDGMENTS
[17] T. McLoughlin, R. S. Laramee, R. Peikert, F. H. Post,
The work described in this paper was carried out in the
and M. Chen. Over two decades of integration-based,
project EarthServer, which has received funding from the
geometric flow visualization. Computer Graphics
European Community’s Seventh Framework Programme (EU
Forum, 29(6):1807–1829, 2010.
FP7-INFRA) under grant agreement no 283610.
[18] Microsoft. Silverlight, 2011.
http://www.microsoft.com/SILVERLIGHT/.
7. REFERENCES [19] N. F. Polys, N. W. John, D. Evestedt, and M. Aratow.
[1] R. Arnaud and M. Barnes. Collada. AK Peters, 2006. Unifying iso standards for medical image presentation.
[2] P. Baumann. Beyond rasters: introducing the new ogc In Workshop on Medical Virtual Env. IEEE, 2010.
web coverage service 2.0. In Proceedings GIS ’10, [20] A. Salga. XB PointStream, 2011.
pages 320–329, New York, NY, USA, 2010. ACM. http://scotland.proximity.on.ca/asalga/demos/freecam/.
[3] J. Behr, Y. Jung, T. Drevensek, and A. Aderhold. [21] T. Schiffer, A. Schiefer, R. Berndt, T. Ullrich,
Dynamic and interactive aspects of X3DOM. In V. Settgast, and D. W. Fellner. Enlightened by the
Proceedings Web3D 2011, pages 81–87, New York, web – a service-oriented architecture for real-time
USA, 2011. ACM Press. photorealistic rendering. In Tagungsband 5. Kongress
[4] J. Behr, Y. Jung, J. Keil, T. Drevensek, P. Eschler, Multimediatechnik Wismar, pages 1–8, 2010.
M. Zöllner, and D. W. Fellner. A scalable architecture [22] C. Schwartz, R. Ruiters, M. Weinmann, and R. Klein.
for the HTML5/ X3D integration model X3DOM. In Webgl-based streaming and presentation framework
S. Spencer, editor, Proceedings Web3D 2010, pages for bidirectional texture functions. In Proceedings
185–193, New York, USA, 2010. ACM Press. VAST 2011, pages 113–120. Eurographics, 2011.
[5] R. Berndt, G. Buchgraber, S. Havemann, V. Settgast, [23] A. Systems. Flash, 2011.
and D. W. Fellner. A publishing workflow for cultural http://www.adobe.com/products/flashplayer.html.
heritage artifacts from 3d-reconstruction to internet [24] C. Thiede, C. Tominski, and H. Schumann.
presentation. In Proceedings of the 3rd intl. conf. on Service-oriented information visualization for smart
digital heritage, pages 166–178, Berlin, 2010. Springer. environments. In 13th Intl. Conf. on Information
[6] R. Bürger and H. Hauser. Visualization of Visualisation, pages 227–234. IEEE, 2009.
multi-variate scientific data. In Eurographics 2007