Research Activities

By utilizing information about the specific type of a material at hand, one can derive domain specific model parameters from a given set of images of a sample. Editing this model can be made much more intuitive compared to general purpose editing techniques. The acquired information about the structure of the material may be further exploited to improve the efficiency of physically based rendering techniques.

The optical reflectance behavior of materials and surfaces is of great importance for the perception of a scene. For example, the specific gloss and the texture of a surface give important hints about properties of the material it is made of. Parametric modeling of this behavior is often difficult or even practically impossible. Therefore, in our group we research techniques for the effective data-driven representation of material appearance based on measurements.

The exploration of combinations of image based geometry and photometric reconstruction methods in our camera dome is currently in the focus of our 3D acquisition research. But we also use standard laser scanner or shape from stereo techniques to capture 3D models. For the processing of the acquired geometry we develop techniques for automatic repair, completion, compression and post-processing.

The creation of digital geometric 3D content requires intuitive and effective surface modeling and editing techniques. In particular, with the advent of 3D laser scanners surface meshes with high resolutions become rapidly available. In order to edit or animate such models, We investigate appropriate mesh editing techniques that allow easy and fast modifications of the scanned surface.

Originally investigated to automatically derive texture maps for surfaces, parameterizations are nowadays an universal tool and prerequisite for many geometry processing application. Our interest lies in energy functionals that capture length, angle and area distortions in mapping from a surface to texture space. Moreover, we explore texture mapping methods for inconsistent or fragmented surfaces as often encountered in practice. We also apply concepts and energy functionals from parameterizations to other areas of computer graphic as e.g. editing or surface deformations.

Spatial information can be most effectively communicated by showing carefully chosen pictures of a scene. This is even more true, if images are taken from a familiar perspective so that the observer can easily match his perspective with the depiction. In contrast to more abstract representations like e.g.\ maps, the mental effort for self localization and orientation is reduced and way finding facilitated. Choice of appropriate viewpoints is, however, non-trivial and requires great skill. Moreover, occlusion significantly restricts the visibility of important features. In our research we explore ways to automatically enhance visibility of important scene features by applying specialized space and surface deformations. Since occlusion provides important depth clues our goal is to minimize occlusion only where it hinders visibility of important features rather than eliminating it altogether.

Along with the more and more important role 3D objects are playing in fields like architecture or engineering, digital libraries tailored to 3D content gain increasing attention. They provide tools for aquisition, automatic metadata generation, retrieval and efficient visualization. Our focus in this research area is on content-based retrieval of 3D objects according to shape and material properties as well as on high-quality visualization. Apart from arbitrary 3D objects, we are especially interested in data related to architecture and cultural heritage.

Realtime visualization of terrain and cities is challenging due to the increasing size and detail of the models.

Since several decades researchers investigate the use of the human hand as a direct input device for enabling different possibilities of human computer interaction. To this end the hand has to be tracked in real-time. Having such a method at hand, suitable interaction techniques are needed for establishing an efficient and intuitive interaction interface. We are working on markerless tracking both human hands as well as on suitable interfaces for complex 3D interaction.

The use of motion capture systems is one of the most important techniques for generating human motion data, nowadays. Modern systems are able to capture complex human motions with a high spatial and temporal resolution. Motion capture data are find applications for motion analysis in sport sciences, biomechanics, medicine and image processing. They are also used for motion synthesis in computer animation in particular. In view of efficient uses of motion capture data we are dealing with questions concerning representation, organization and reduction of these.

We are concerned with all aspect of hair modeling: the generation of hair styles, physics-based simulation of strands and wisps, and photorealistic and efficient rendering techniques. Also the inverse problems are a central topic of our research.

We are concorned with symbolic methods for the analysis of deterministic simulations models.Our main focus is the parametric analysis of ordinary differential equations. Using quantifier elimination over the reals we perform symbolic bifurcation analysis and apply these techniques in various contexts.