Changes in the scene (light sources, geometry, daytime …) lead to an immediate need for a recalculation of the illumination result. By using an iteratively improving approach on modern GPU hardware, which converges to the physically correct solution after several hundred frames, scene interactions are still made possible during these costly computations – even in cases of a large number of physically complex light sources.

The developed system will make it possible to visualize the simulation results on both regular monitors as well as on 3D stereo setups with multiple screens, providing an immersive impression of the illuminated buildings. By applying an interactive graphical user interface, it will be both possible to modify light source settings (like position, type, color and intensity) and scene geometry attributes (like textures or object positions/orientation), offering on-the-fly changes in scene and light simulation configuration.

Publications

Current research topics include fast, physically plausible illumination solutions, handling of very large geospatial data sets, hard real-time tracking and graphics processing for broadcast video, and algorithms for automatic acquisition, simplification and 3d model generation.

Recent Work and Achievements:

O-Snap is an optimization-based modeling and reconstruction framework for architectural structures: Based on novel automatic point cloud and adjacency analysis techniques, scanned buildings are reconstructed as coarse, low-polygonal meshes. By introducing an intuitive, sketch-based user interface, holes and errors in the data can be fixed with a few simple clicks, while the interactive optimization guarantees the polygons to fit the point cloud data in the best possible way. This project, started in the "Semantic Modeling and Acquisition Group", acts as a starting point for further research in the area of semantic scene understanding and 3D reconstruction - for example in the upcoming REPLICATE project. A first publication will shortly appear in the journal "ACM Transaction on Graphics (TOG)" - visit the publication website here.

Recent work and research on the use of physically accurate soft shadows in real-time applications has proven to be valuable in various industry-related projects. Especially the HILITE project, which focuses on the calculation of global illumination using a shadow mapping-based approach, has benefited from the gained knowledge. Publications: Fast Percentage Closer Soft Shadows using Temporal Coherence Fast Accurate Soft Shadows with Adaptive Light Source Sampling Fast Soft Shadows with Temporal Coherence Real-Time Soft Shadows Using Temporal Coherence

Such information may come from a wide variety of sources, such as government institutions or private service providers.

This project researched concepts and technologies that shall facilitate the communication, collection and processing of this information, especially in a geospatial context. Two of the key requirements were an open design to accommodate a large variety of data with a straightforward workflow for integration within the system, and a shared workspace for all participating agencies.

To demonstrate and validate the candidate technologies, a prototype demonstrator system was built and used in two simulated scenarios. The feedback from these full-day events was then used to refine the design of the system.

The system was designed as a web based architecture for easy deployment and shared access. Shared messaging, task management, full-text searchable document storage and other collaboration functions supported closed-group communication on sensitive topics; a powerful map display and visualization module became the main component for the visualization and exploration of task-relevant geospatial information.

A simple administration user interface allows non-experts to enrich the database using standard geospatial and office file formats (ESRI Shapefile, PDF, Word and Excel documents, etc.), select subsets of the available information for display and perform other routine maintenance tasks.

Several design iterations led to a system that utilizes an open, document-oriented backend database architecture based on MongoDB, which can easily accommodate arbitrarily structured information (even within single collections), allows easy updating and extension of stored entities, and supports replication and clustering for a scalable and reliable system. To allow highly efficient geospatial queries, a PostGIS front-end database caches the stored entities and natively supports a wide range of geospatial queries and calculations.

Thanks to recent advances in HTML5 and JavaScript performance, significant parts of the system could be implemented to run inside the web browser, including a flexible map display component based on OpenLayers, geospatial queries and interactive visualization overlays. Therefore, the system requires essentially no client-side installation (except for a recent browser), facilitating deployment in secure environments, and is highly scalable.

In addition, GIS metadata can also be used to populate virtual environments with autonomous vehicles, control geometry creation and provide valuable additional information (such as noise levels in highway simulations) to the viewer. The prototype viewer has already been used for various public audits and presentations, and has been generally recognized as a valuable tool in these situations.

Publications

Piringer, H, Buchetics, M, and Benedik, R (2012).
AlVis: Situation Awareness in the Surveillance of Road Tunnels
in IEEE Transactions on Visualization and Computer Graphics / Proceedings of IEEE VAST 2012.

Mission: To support time-critical decision making using visual simulation control.

Fluid simulation tools are capable of predicting natural processes and can be employed for the assistance in the human decision making process. Existing solutions lack a number of important features needed for a feasible support system. Very important is the usability of the simulation tools by people without special fluid simulation expertise. Moreover, with the capabilities of graphics hardware clusters, everybody will have access to affordable supercomputing power on their desktop. Until now, little work has been done to use this power for knowledge generation via simulation steering. In this project we develop a novel computational steering system that uses visualization in every aspect of the problem solving and provides effective feedback via an intuitive interface.

Among the application areas of the system we consider the industrial design of components where rapid prototyping is required. Using a fast and intuitive system will help to evaluate whether a certain concept is promising during the design phase. Another important application is the assistance in emergency situations that are caused by natural disasters such as floods, where safety and damage limitation depend on fast decisions. The system can be used to test actions to be taken during a flood event . Users interact with sketch-based steering monitors to quickly plan and design a breach closure, to name an example. The gained knowledge supports the creation of flood management plans. Finally, our vision is that, even under time-critical circumstances, emergency personnel on-site will be able to analyze the imminent situation quickly to choose the best response strategy.

World Lines are introduced as an intuitive representation of multiple simulation runs. This concept allows to create, manage and compare alternative scenarios with the goal to understand the complex interplay between simulation input parameters and simulation outcome. The underlying data-flow network enables a flexible integration of simulation, visualization and steering components into a single application. This way, we can adapt several existing simulation modules and integrate them. Visdom comprises a client-server architecture to prepare for the future vision of decision support on-site using mobile devices. The compute-intensive parts are written as server modules controlled over the web. Visdom exploits fast GPU technology for high-performance simulation, analysis and visualization.

Links

Visdom Screenshots

Intuitive design of a breach closure using semantic steering monitors (top views) that are linked to World Lines (bottom view). The interactive World Lines view represents alternative choices (multiple related simulation runs) as a set of causally connected tracks.

Comparative visual analysis of a loaded dam break data set.

Visdom is a modular system allowing the combination of multiple steering monitors or views for comparative analysis of alternative scenarios.

Visual analysis has become a scientifically and economically important field. Multiple publications, a powerful software framework, and a growing number of related industry projects show that visual analysis is a key competence of VRVis. Building on this competence and technology, the strategic research project IVAN has two high-level goals: First, it seeks to address challenging research topics which further increase the visibility of visual analysis research of VRVis. Second, it will ensure the high degree of innovation of the visual analysis-related software technology of VRVis for the next few years.

IVAN addresses four concrete topics of research which have evolved as strategically relevant in discussions with both science and industry.

1. The first topic addresses the exploration of continuous parameter spaces using multivariate prediction. Based on statistical methods, we envision to enable a local analysis of continuous, sampled parameter spaces and a prediction of quantitative target values in real-time. Novel visualization techniques will provide guidance for an efficient navigation to interesting regions of continuous parameter spaces and a local sensitivity analysis with respect to multiple parameters (see also the image). Another aspect concerns the visualization of the inherent uncertainty of predictions considering the different sources of uncertainty for different prediction methods.
2. The second topic focuses on the generalization of data derivation as a powerful approach to coordinate multiple views. Today, data derivation is typically a static part of pre-processing. On the other hand, certain types of data derivation (e.g., interactive selection by brushing) have tightly been integrated in the analysis but are not general with respect to interaction and visualization concepts. We intend to describe a general model for interactive data derivation that seeks to combine two goals: 1) the specification of data derivation should be tightly integrated in the analysis process, and 2) the results should be handled as flexibly as possible. Based on an implementation of the model within visplore, we envision many applications that could benefit from a generalized approach, including similarity-based analysis, interactive data editing, and advanced aggregations of time-dependent data.
3. The third topic is dedicated to a comparative visualization of many categories. Many current visualization approaches for analyzing categorical data rely on side-by-side comparison in small-multiple visualizations. While a small-multiple layout will also be the starting point of our approach, we intend to exceed the capabilities of current approaches in two important aspects: first, the envisioned approach should be tightly coupled to all other views, including multiple instances of itself. Our second goal is to explicitly visualize the difference of each plot with respect to a reference plot for common visualization types like bar charts, scatter plots, time series, and box plots. We expect such comparisons to be more precise than side-by-side comparisons in the case of many categories.
4. The fourth topic deals with the visual analysis of many relational data tables. A relational data model is widely used in databases and data warehouses. On the other hand, many visualizations are limited to represent the data in terms of the raw data records that constitute one imported data table at a time. We intend to enable a simultaneous analysis of multiple relational data tables in visplore by linking views that refer to different tables. Another goal is to provide an overview of a database comprising many relational data tables, and to support to defining new pivot tables in an ad-hoc manner.

The work on these topics includes the conceptualization, prototype implementation within the system visplore, and the evaluation of results. Moreover, IVAN involves the scientific dissemination of results and scientific networking as well as the supervision of students.

Our galleries are full of wonderful paintings – some so famous that nearly everyone knows about them. Unfortunately, blind and visually impaired people are mostly excluded from the world of visual arts. Some museums offer special guided tours describing selected paintings verbally. However, it is extremely difficult to crate a mental image from acoustic impressions only. State of the art tactile diagrams (a stylized version of the painting, mostly line drawings embossed on paper) help getting an overview, but allow only a very simplified view. On the other side, hand-crafted bas reliefs are very detailed and good to read, but require skilled sculptors in the fully manual creation process. We wanted to create a process that doesn't require any manual skills, and allows the creation of different tactile representations of suitable complexity, while being faithful to the original artworks.

At the current stage of the project, we developed a computer-assisted process, that converts images into three different tactile media of increasing complexity. This process consists of three consecutive stages. At the end of each stage, the data for the production of one tactile media is available.

In the first stage, important structures are identified, resulting in line drawings that separates the images into semantically meaningful entities. These drawings, vectorized and augmented with different fill patterns can be used to produce tactile diagrams.

The second stage adds in depth that is an important part in most figurative paintings. The human eye is trained in decoding this hidden third dimension, while the tactile sense is used to getting three-dimensional input directly. A custom built, intuitive user interface allows the artist to quickly annotate depth relations, and the software automatically assigns an appropriate depth to each object in the painting. The resulting discrete depth map is converted into layers to be laser-cut out of flexible plastic sheets and assembled on top of each other to form a "layered depth diagram".

In the third stage, texture information is extracted from the original paintings using customizable filter stages (similar to processing stages in the human eye) and superimposed on the layered depth diagrams. With an interactive, three-dimensional preview software, the filter settings can be optimized, perspective effects can be enabled, and correction-layers from external software can be mixed in. The resulting "textured reliefs" can be manufactured with computer-controlled CNC milling machines. It takes several hours, until the fine milling tools carve out all the fine detail. Custom software to create the machine codes had to be developed, since no commercial CAM software was able to process this highly detailed data. From a negative cast, mutliple copies can be mold. Several materials have been tested, until a durable, stain-resistant and pleasant material was found.

The project was performed in cooperation with Kunsthistorisches Museum Vienna and was funded by "KulturKontakt Austria im Auftrag des BMUKK". Four textured reliefs for three paintings with accompanying descriptions in braille are currently available, and special guided tours are offered.

Videos

• http://www.youtube.com/watch?v=SkbIqTrYSUk

Links

Publications/Talks

• A. Reichinger, M. Neumüller, F. Rist, S. Maierhofer, W. Purgathofer. "Computer-Aided Design of Tactile Models - Taxonomy and Case Studies". In Miesenberger, K., Karshmer, A., Penaz, P., Zagler, W., eds.: Computers Helping People with Special Needs. 7383 Volume of Lecture Notes in Computer Science. Springer Berlin / Heidelberg (2012), pp. 497-504. PDF, Talk
• A. Reichinger, S. Maierhofer, and W. Purgathofer. High-Quality Tactile Paintings. ACM J. Comput. Cult. Herit. 4, 2, Article 5 (November 2011), 13 pages. DOI = http://doi.acm.org/10.1145/2037820.2037822. PDF
• A. Reichinger, S. Maierhofer, W. Purgathofer, High-Quality Tactile Paintings. In Eurographics 2011 - Areas Papers, April 2011, pp. 1-8 (PDF).
• A. Reichinger, Gallery Paintings for Blind and Visually Impaired People, Talk at SpaceX - An Exchange Forum on Information Design for Visually Impaired People, Vienna, October 25-26, 2010.

Awards

• Juryauszeichnung Multimedia und e-Business Staatspreis 2010/11: Innovationspreis.
• Mercur'11 der Innovationspreis der Wirtschaftskammer Wien: Nominiert in der Kategorie Kreativität

Presentations

• Wiener Forschungsfest 2010, 18.-20.9.2010.
• eday 2011, 3.3.2011.
• OCG Horizonte, 19.6.2012.
• OCG Talk am Campus, Hagenberg, 22.10.2012.

Press, Reports, TV

Examples

Raffael, "Madonna im Grünen" (Madonna of the Meadow), 1505 or 1506.

KHM image database

Line diagram and Layered Depth Diagram.

Textured Relief.

Textured Relief of closeup (background top left).

Jean Fouquet, "Der ferraresische Hofnarr Gonella" (Portrait of the Ferrara Court Jester Gonella), around 1445.

KHM image database

Textured Relief

Textured Relief.

VILMA consists of a dynamic vision -- sensor setup (e.g. digital cameras, laser scan, thermal imaging), aiming at a geo-referenced data acquisition speed of 1 km / hour in minimum and a mapping resolution of better than 1 mm specifically at on-line selected regions of interest such as cracks or water ingress in tunnels. During data acquisition, on-site visualisation is a major development issue to efficiently guide the operator and immediately assess data usability and quality. The different sensor data is co-registered, geo-referenced, and integrated into a hierarchical surface representation for efficient access to layers of different resolution, thematic content, stage of data processing, and spatial as well as temporal information such as data acquisition time, deformation dynamics, or construction progress. Visualization issues cover the user-friendly real-time random 2D access to surface layers, and a typical set of 3D visualization functions that fulfils the operational and quality needs of the construction and maintenance site. To demonstrate the sustainable VILMA application potential some highly relevant quality management and cost aspects are addressed such as crack detection and evaluation, requirements from geology, and frequent monitoring. The system is scalable in terms of sensor setup, processing complexity and visualization hardware performance. The integration of external sensor data like handheld camera images, the practicability of each system component, assessed by the user, and the awareness to up-to-date 2D & 3D visualization developments are important development drivers.

VILMA integrates challenging aspects of vision sensor design, 3D reconstruction, recognition, and visualization. The proposing partners unify application awareness, long-term experience in the related scientific topics and proven success of recent joint developments.

At present the European Industry recognised their obligation to reconsider risk and safety policies, having a more competitive industry and more risk informed and innovation accepting society in vision. Therefore the large collaborative project IRIS is proposed to identify, quantify and mitigate existing and emerging risks to create societal cost-benefits, to increase industrial safety and to reduce impact on human health and environment.

The project is led and driven by industry to consolidate and generate knowledge and technologies which enable the integration of new safety concepts related to technical, human, organizational and cultural aspects. The partnership represents over 1 million workers. The project integrates all aspects of industrial safety with some priority on saving human lives prior cost reductions and is particular underpinning relevant EU policies.

Recent developments in the target domains of this project show a clear transition from steady to unsteady flow scenarios.  Accordingly, we have to see that the traditionally proven approaches do not apply any more and that a conceptual change in the methodology of visual analysis is necessary.  Topology-based methods which account for the complete dynamic behaviour of flow fields are strongly needed but do not exist.  Steps toward this goal have been done from several sides, delivering promising but yet only partial results.

It is the objective of this project to research a new segmentation method for unsteady flows that has the elegance and specificity of (steady) VFT, but which provides correct results for unsteady flows as well.  This project aims at the formulation of a sound theoretical mechanism to describe structural features in time-dependent flow.  Similar to the case of steady flow, were topology has proven its usefulness in many years, it is straight-forward to expect that the new approach will also establish its important role in the analysis and discussion of time-dependent flow scenarios.  As part of a successful project, concrete algorithms to extract and visualize the topological structures are derived from the new mechanism.  Implementations of them will allow studying the usefulness on a number of real-life flow data from different areas of application.

MIIVIS will be primarily linked to the ENGVIS project, benefiting from all visualization and analysis approaches dealing with large and 3D data. Especially the linking of techniques from scientific visualization with information visualization will also play an important role in future applications dealing with large volumetric data from numerous sources.