Marcel Lancelle - Publications

This thesis covers research on new and alternative ways of interaction with computers. Virtual Reality and multi touch setups are discussed with a focus on three dimensional rendering and photographic applications in the field of Computer Graphics. Virtual Reality (VR) and Virtual Environments (VE) were once thought to be the future interface to computers. However, a lot of problems prevent an everyday use. This work shows solutions to some of the problems and discusses remaining issues.
Hardware for Virtual Reality is diverse and many new devices are still being developed. An overview on historic and current devices and VE setups is given and our setups are described. The DAVE, an immersive projection room, and the HEyeWall Graz, a large high resolution display with multi touch input are presented. Available processing power and in some parts rapidly decreasing prices lead to a continuous change of the best choice of hardware. A major influence of this choice is the application. VR and multi touch setups often require sensing or tracking the user, optical tracking being a common choice. Hardware and software of an optical 3D marker tracking and an optical multi touch system are explained.
The Davelib, a software framework for rendering 3D models in Virtual Environments is presented. It allows to easily port existing 3D applications to immersive setups with stereoscopic rendering and head tracking. Display calibration and rendering issues that are special to VR setups are explained. User interfaces for navigation and manipulation are described, focusing on interaction techniques for the DAVE and for multi touch screens. Intuitive methods are shown that are easy to learn and use, even for computer illiterates. Exemplary applications demonstrate the potential of immersive and non-immersive setups, showing which applications can most benefit from Virtual Environments. Also, some image processing applications in the area of computational photography are explained, that help to better depict the captured scene.

Today’s super zoom cameras offer a large optical zoom range of over 30x. It is easy to take a wide angle photograph of the scene together with a few zoomed in high resolution crops. Only little work has been done to appropriately display the high resolution photo as an inset. Usually, to hide the resolution transition, alpha blending is used. Visible transition boundaries or ghosting artifacts may result. In this paper we introduce a different, novel approach to overcome these problems. Across the transition, we gradually attenuate the maximum image frequency. We achieve this with a Gaussian blur with an exponentially increasing standard deviation.

We address artifacts at corners in soft edge blend masks for tiled projector arrays. We compare existing and novel modifications of the commonly used weighting function and analyze the first order discontinuities of the resulting blend masks. In practice, e.g. when the projector lamps are not equally bright or with rear projection screens, these discontinuities may lead to visible artifacts. By using first order continuous weighting functions, we achieve significantly smoother results compared to commonly used blend masks.

Existing visual surveillance systems typically require that human operators observe video streams from different cameras, which becomes infeasible if the number of observed cameras is ever increasing. In this paper, we present a new surveillance system that combines automatic video analysis (i.e., single person tracking and crowd analysis) and interactive visualization. Our novel visualization takes advantage of a high resolution display and given 3D information to focus the operator’s attention to interesting/critical areas of the observed area. This is realized by embedding the results of automatic scene analysis techniques into the visualization. By providing different visualization modes, the user can easily switch between the different modes and can select the mode which provides most information. The system is demonstrated for a real setup on a university campus.

Camera-based object detection and tracking are image processing tasks that typically do not take 3D information into account. Spatial relations, however, are sometimes crucial to judge the correctness or importance of detection and tracking results. Especially in applications with a large number of image processing tasks running in parallel, traditional methods of presenting detection results do not scale. In such cases it can be very useful to transform the detection results back into their common 3D space. We present a computer graphics system that is capable of showing a large number of detection results in real-time, using different levels of abstraction, on various hardware configurations. As example application we demonstrate our system with a surveillance task involving eight cameras.

The DAVE is an immersive projection environment, a foursided CAVE. DAVE stands for ’definitely affordable virtual environment’. ’Affordable’ means that by mostly using standard hardware components we can greatly reduce costs compared to other commercial systems. We show the hardware setup and some applications in the accompaning video. In 2005 we buildt a new version of our DAVE at the University of Technology in Graz, Austria. Room restrictions motivated a new compact design to optimally use the available space. The back projection material with a custom shape is streched to the wooden frame to provide a flat surface without ripples.

We present several novel ways of interaction and navigation in virtual worlds. Using the optical tracking system of our four-sided Definitely Affordable Virtual Environment (DAVE), we designed and implemented navigation and movement controls using the user’s gestures and postures. Our techniques are more natural and intuitive than a standard 3D joystick-based approach, which compromises the immersion’s impact.

We describe first steps towards a suite of tools for CH professionals to set up and run digital exhibitions of cultural 3D artifacts in museums. Both the authoring and the presentation views shall finally be as easy to use as, e.g., Microsoft Powerpoint. But instead of separated slides our tool uses pre-defined 3D scenes, called "layouts", containing geometric objects acting as placeholders, called "drop targets". They can be replaced quite easily, in a drag-and-drop fashion, by digitized 3D models, and also by text and images, to customize and adapt a digital exhibition to the style of the real museum. Furthermore, the tool set contains easy-to-use tools for the rapid 3D modeling of simple geometry and for the alignment of given models to a common coordinate system.
The technical innovation is that the tool set is not a monolithic application. Instead it is completely based on scripted designs, using the OpenSG scene graph engine and the GML scripting language. This makes it extremely flexible: Anybody capable of drag-and-drop can design 3D exhibitions. Anybody capable of GML scripting can create new designs. And finally, we claim that the presentation setup of our designs is 'grandparent-compliant', meaning that it permits to the public audience the detailed inspection of beautiful cultural 3D objects without getting lost or feeling uncomfortable.

Addressing the typically time consuming adjustment of projector equipment in VR installations we propose an easy to implement projector calibration method that effectively corrects images projected onto planar surfaces and which does not require any additional hardware. For hardware accelerated 3D applications only the projection matrix has to be modified slightly thus there is no performance impact and existing applications can be adopted easily.

Leapfrog 3D is 3D visualization software for geological modeling focusing on the mining domain. The tool allows users to filter, analyze and visualize different types of underground measurements (drill hole, seismic data, etc.). The software provides a much faster way of generating possible raw grade morphological models compared to the long explicit modeling sessions needed with current classical tools. By using implicit modeling and a proprietary interpolating algorithm (fastRBF^TM), geologists can simply introduce an interpreted morpholocical model, and the software will converge rapidly to a new meshing solution (Figure 1).
This report is a response to the company request to explore new interface tools that can assist with product presentation and also provide easy navigation with a better depth perception. Moreover, the company is also interested in exploring the possibilities of a Post-WIMP¹ Interface. We present in this document requirements identified during our observations and discussions, following by a review of current state of the art interface techniques (section 3) and a selected of possible solution ideas (section 4).

Many different projects exist that are more or less related to large public interactive displays. Most of them exist for research purposes. The following text describes different issues, ideas and existing solutions.

This research covers issues of automatic generation and visualization of a 3D city model using existing digital data. Air photos, LIDAR point clouds and soil usage and cadastral maps are used as data sources for the automated data fusion in a sample implementation. Combining existing techniques and new ideas, the developed framework enables automatic information enrichment and generation of plausible details.
Data conversion, model generation and real time visualization are achieved using a standard PC for the whole area of 200km². Real world data was used including all the practical problems involved.

Das Thema der 3D-Stadtmodelle ist sehr vielfältig und interdisziplinär. Dieses Kapitel gibt einen Überblick über den behandelten Stoff und die Ziele dieser Diplomarbeit.
Neben theoretischen Überlegungen sollte auch die konkrete Realisierung einzelner Aspekte unter den gegebenen Einschränkungen erfolgen. Ein schnelles Ergebnis ohne zusätzliche Kosten erforderte die automatische Verarbeitung verfügbarer bereits vorhandener digitaler Daten. Dadurch wurde es auch möglich, die Daten für eine ganze Stadt zu verarbeiten. Mit Hilfe der Implementierung können Forschung und Praxis verknüpft werden und zahlreiche weitere Forschungsprojekte bieten sich an, die auf dieser Grundlage aufbauen können. Es stand kein Komplett-System zur Verfügung, auf das aufgebaut werden konnte, so dass von Grund auf neu angefangen und vieles neu implementiert werden musste. Dadurch hielten sich zwar die Einschränkungen der externen Komponenten in Grenzen, doch war nur Zeit für einfache Realisierungen; es gibt für alle Teilbereiche der Erzeugung und Visualisierung bereits jeweils bessere Implementierungen. Der entstandene "CityModelViewer" ist dennoch bereits ein nützliches Tool, das diverse Arten von Daten aus ganz unterschiedlichen Quellen nutzen und bereits auf einem durchschnittlich ausgestatteten PC interaktiv visualisieren kann.
Mit Hilfe von Daten der Stadt Braunschweig wurden nach und nach immer mehr Programmteile entwickelt. So wurden hauptsächlich die für diese Stadt relevanten Probleme behandelt. Die konkrete Bearbeitung von Problemen ist ohne vorhandene realistische Daten kaum sinnvoll und wurde daher wenig berücksichtigt. Für Braunschweig existierte noch gar kein digitales 3D-Stadtmodell, so dass dies eine gute Untersuchung für später geplante professionelle Projekte darstellt. So können auch die Mitarbeiter der Stadt schon jetzt erste praktische Erfahrungen auf diesem Gebiet sammeln. Eine Nutzung außerhalb der genannten Bereiche, z.B. im privaten Umfeld, ist leider aus rechtlichen Gründen wegen der benutzten Daten nur sehr eingeschränkt möglich.

Die Erstellung von 3D Modellen ist ein wichtiger Schritt in der Computergraphik. Mit heutigen Modellierungswerkzeugen ist dies für manche Objekte einfach zu erfüllen, doch gibt es auch viele Objekte, bei denen eine Modellierung sehr aufwendig wäre. 3D-Scanner helfen hier weiter, sofern das Objekt schon existiert und verfügbar ist. Bei jedem der vielen Verfahren des 3D Scannens gibt es Einschränkungen bezüglich der Größe, Form, Konsistenz oder anderen Materialeigenschaften des Objektes. Doch das 3D-Scannen wird häufig einfach aus finanziellen Gründen nicht eingesetzt.
Das Ziel dieser Studienarbeit war die Konstruktion einer einfachen und preisgünstigen Hardware sowie die Erstellung der Software. Das Verfahren "shape from silhouette" sollte benutzt und für automatisierte Aufnahmen ein Drehteller konstruiert werden. Dabei wird das Objekt aus verschiedenen Blickwinkeln vor einem einfarbigen Hintergrund fotografiert und die Anteile jedes Fotos außerhalb des Objektes aus einem soliden Block ausgestanzt. Dadurch wird allerdings nur die konvexe Hülle des Objekts rekonstruiert, was der größte Nachteil dieses Ansatzes ist.
Für die automatische Segmentierung der Bilder darf das Objekt auch keine Farben ähnlich der Hintergrundfarbe enthalten bzw. diese stark reflektieren oder transparent durchscheinend sein.
Ein Vorteil ist die Möglichkeit der manuellen Korrektur der Segmentierung, was prinzipiell das Scannen von Objekten mit beliebiger Oberfläche erlaubt. Ebenfalls vorteilhaft sind die geringen Hardwarekosten. Mit einiger manueller Arbeit ist sogar eine grobe Rekonstruktion eines Objekts mit Hilfe von normalen, unkalibrierten Fotos möglich.