PhD Stanford, 1994
My work in the field of computer vision covers both basic research
issues and a range of new applications. I have worked on a number of issues in early
vision, including image restoration, visual correspondence, and motion-based tracking. My
recent work with Y. Boykov and O. Veksler provides novel algorithms for a fundamental
class of vision problems whose previous solutions required exponential time. Our results
rely on the application of recent results from graph theory, which can provide efficient
algorithms that generate provably good answers.
I am also exploring a number of different
applications of computer vision, with an emphasis on automating visual tasks. One of the
most important applications is in the area of content-based access to image databases.
While large collections of digital imagery are becoming commonplace, the tools available
for accessing image databases are still quite primitive. My past work with G. Pass and
with J. Huang et al. focused on this problem for collections of static images. In ongoing
research, V. Kettnaker and I are addressing the issue for video, where we hope to exploit
some insights about the role of contextual information in the human visual system. I am
also working with Microsoft on automating some visual tasks that occur in the development
of programs with graphical user interfaces.
Program Committee, ACM Conf. on Multimedia.
A combinatorial approach to early vision.
Microsoft Research, Dec. 1997.
___. Computer Science, Dartmouth, April 1998.
Automatic hierarchical color image classification.
Proc. ACM Conf. Multimedia, Bristol, England, Aug. 1998 (with J. Huang and R.
Markov random fields with efficient
approximations. Proc. IEEE Computer Vision and Pattern Recognition Conf., Santa
Barbara, California, June 1998 (with Y. Boykov and O. Veksler).