standing Architecture, we are actively exploring computational architectures suited to the types of processing inherent in image understanding, explicitly taking into account the set of algorithms presented in this chapter and the bandwidth requirements of the communication channels between the different levels. When a physical slice of the architecture is constructed it will be used to run real-time vision interpretation experiments.
The authors wish to thank the members of the VISIONS research group, both past and present, for their contributions to a long-term dream. Since the beginning, there have been moments where particular individuals have provided significant leadership in the maturation of the ideas.
We wish to specifically acknowledge several individuals who have made major contributions to our 12-year effort. Tom Williams was there at the inception of the VISIONS project and contributed greatly to the design of the system through the first 6 years. Ralf Kohler designed the VISION software development environment that has supported empirical development in a flexible manner for many years. Joey Griffith has had his fingers in almost every aspect of the project for the last 6 years. George Reynolds has taken on a significant leadership role in the development of many aspects of intermediate processing. We also wish to acknowledge the leadership provided by Daryl Lawton and P. Anandan in developing our motion group.
In addition, a number of other people have made their presence felt including Terry Weymouth, John Lowrance, Frank Glazer, Paul Nagin, Bryant York, Gilad Adiv, Charlie Kohl, Len Wesley, Cesare Parma, and current members Bruce Draper, Rob Belknap, Bob Collins, Michael Boldt, Brian Burns, Ross Beveridge, Nancy Lehrer, Seraj Bharwani, Val Cohen, Jim Burrill, Bob Heller, Mark Snyder, Igor Pavlin, Rich Weiss, Les Kitchen, Chip Weems, and Steve Levitan.
We also wish to thank Janet Turnbull for her continuing support over the years, and Laurie Waskiewicz for her perserverance and good humor during the production of this chapter.
This work has been supported by the Air Force Office of Scientific Research under grant F49620-83-C-0099, by the Defense Advanced Research Projects Agency under contracts N00014-82-K-0464 and DACA76-85-C-0008, and by the National Science Foundation under grant DCR-8318776.