Abstract
An optical system that was constructed to implement the Marr-Poggio stereo algorithm is described. This particular problem was chosen because it represented an interesting framework in which to study some of the broader issues of the implementation of optical neural networks. The problem of handling bipolar quantities is addressed. Some conclusions are drawn concerning the technology most appropriate to the implementation of nodes in a general optical neurocomputer.
Original language | English (US) |
---|---|
Pages | iii/599-606 |
State | Published - 1987 |
ASJC Scopus subject areas
- Engineering(all)