Statistical Geometry Processing (Former Research Group)

Vision and Research Strategy

The objective of this research group is to investigate statistical data analysis techniques for geometric data sets. Informally described, the long-term goal of this research direction is to devise techniques to “understand” geometric data sets and to use the acquired knowledge to aid geometric modeling and geometry processing. There are two main motivations for addressing these questions:

  • Practical relevance: The acquisition of real world objects as 3D models has become an indispensable tool in a large number of applications, such as geoinformation systems (think “Google Earth”) or architectural planning. However, 3D scanning has severe limitations: all practical 3D acquisition techniques suffer from noise and outlier artifacts, as well as incomplete acquisition due to occlusion. Statistical reconstruction allows us to improve the model quality without additional on-site scanning, which can save a lot of money. In creative application areas, such as movie making or content creation for computer games, the costs for content creation are nowadays the dominating factor limiting the applicability of 3D computer graphics techniques. Here, data analysis algorithms can help to understand the structure of the data and aid the user in creating variations of models in their data base that fit their needs.
  • Philosophical insight: The second aspect that motivates this research direction is the nature of the problem itself. Before we can find algorithms that “understand the structure” of geometric data sets, we need to come up with a formal definition of what constitutes this “structure” and makes it objectively identifiable. Ultimately, this leads to the question of how human perception actually works and what is necessary to formally constitute cognitive models: how much of what we call structure can be covered by simple mathematical formalisms and how much human knowledge (derived from millennia of cultural history) is necessary? In this respect, our research area is very similar to the field of computer vision, which tries to “understand” images. However, by working on 3D geometry rather than image data, we avoid many of the very hard inverse problems of reconstructing information from images, and thus may be able to gain new insights from this different perspective.

Composition of Group

The Statistical Geometry Processing Group headed by Michael Wand was established in June 2008. The group currently consists of five PhD students (Martin Bokeloh, Silke Jansen, Jens Kerber, Art Tevs, and Martin Sunkel) and one postdoc (Ruxandra Lasowski). Of these, Jens Kerber and Ruxandra Lasowski are funded by the Cluster.

Research Topics and Achievements

Projects and Collaborations

Dr. Michael Wand

Dr. Michael Wand

Michael Wand headed the Independent Research Group Statistical Geometry Processing from June 2008 until Mai 2013.

Fon: +49 681 9325 4008