Warning:
This page is
provided for historical and archival purposes
only. While the seminar dates are correct, we offer no
guarantee of informational accuracy or link
validity. Contact information for the speakers, hosts and
seminar committee are certainly out of date.
One of the most difficult and basic problems in computer vision is segmentation: grouping sets of pixels together in a meaningful way. Early approaches to segmentation focused only on image features such as color or intensity without regard for their physical interpretation. At CMU in the mid-80's the work of Klinker, Shafer, and Kanade showed that by using physical models of the interaction of light and matter to understand images we could not only obtain better segmentations, but also a better understanding of the scene.
For the past four years, Steve Shafer and myself have worked on a general framework for segmentation using physical models of image formation. This framework proposes multiple physical hypotheses for simple image regions to represent the ambiguity present in an image. Reasoning about these hypotheses allows us to define the space of potential segmentations.
A consequence of this framework was a proposal for a new approach to the segmentation of complex scenes into regions corresponding to coherent surfaces. In this talk, I will present an implementation of this new approach and show example segmentations of scenes containing multi-colored piece-wise uniform objects. By using this new approach we are able to intelligently segment scenes with objects of greater complexity than previous physics-based segmentation algorithms. The results show that by using general simple image regions to represent the ambiguity present in an image. Reasoning about these hypotheses allows us to define the space of potential segmentations.
Bruce is a 4th year Robotics Ph.D. student and a member of almost all of the Viking Death Rat sports teams. Before arriving (late) at CMU, he passed through Swarthmore College and Cambridge University in a vain attempt to pad his resume with degrees. His primary area of research is physics-based vision, focusing on the intelligent segmentation of color images containing multi-colored objects. Bruce's advisor is Steve Shafer, who now passes along his wisdom long-distance from Microsoft Research in Seattle.