IMA:13th Conference on The Mathematics of Surfaces

Abstract
Morphable models constitute a unifying framework for the analysis and synthesis of images. In the field of Computer Graphics, they are applied to model photo-realistic face images; in the domain of Computer Vision, they are used in face recognition applications compensating variations across pose, illumination and facial expressions. Morphable face models draw on prior knowledge of human faces in the form of a general face model, learned from examples of other faces. By exploiting the correspondences between all examples, these models introduce a vector space structure on the examples that allows to synthesize novel photo-realistic images. Image analysis can be performed by fitting such a flexible model to novel images. Then, the model parameters yielding the optimal reconstruction are used to code or analyze the face depicted. In this talk, I start with a quick review on morphable face models and will discuss some of its limitations for face recognition applications based on skin detail analysis. In a second part I will report on current work on building morphable models of human bones and skulls and their application in medical data analysis for image segmentation and the probabilistic modeling of traumatized body parts.

Thomas Vetter studied mathematics and physics and received the Ph.D. degree in biophysics from the University of Ulm, Germany. As a Postdoctoral Researcher at the Center for Biological and Computational Learning, Massachusetts Institute of Technology, Cambridge, he started his research on computer vision. In 1993, he moved to the Max-Planck-Institute, Tübingen, Germany, and, in 1999, he became a Professor of computer graphics at the University of Freiburg, Germany.

Since 2002, he has been a Professor of applied computer science at the University of Basel, Switzerland. His current research is on image understanding, graphics, and automated model building.

Back