Parallel Algorithms for Computer Vision
Abstract
Much work during the past year has focused on building our Vision Machine system. The Vision Machine is a testbed for the research on parallel vision algorithms and their integration. The system consists of an input device- -a movable two camera Eye-Head system with six degrees of freedom--and the 16K Connection Machine (CM-1). We have concentrated on implementing and testing early vision algorithms, and on developing a new sophisticated techniques for their integration. The output of the integration stage will be used for navigation and recognition tasks. From August 31, 1986 to August 31, 1987, we have been using the Connection Machine delivered on July 31, 1986 by Thinking Machines corporation. We have developed and tested a substantial body of vision software on the machine. We have also nearly completed, well ahead of schedule, the development of an integrated Vision Machine that includes several early vision algorithms, and integration stage of middle vision. As outlined in our original proposal, we have begun to explore parallel algorithms at the higher level of recognition. We have also studied the performance of alternative, nonconventional architectures for navigation, and worked on the difficult issue of alternative parallel languages for the Connection Machine, in addition to LISP and C. The body of this report gives an overview of the results of our research during the second twelve month of funding.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1988
- Accession Number
- ADA203947
Entities
People
- James Little
- Tomaso Poggio
Organizations
- Massachusetts Institute of Technology