Abstract
The problem of recognizing patterns in photographs is considered. It is assumed that the patterns to be classified are exactly known, and that the photograph is sampled and quantized to form a binary vector. The optimum classification procedure for the general case involves nonlinear data processing on this vector. In two important special cases, however, this optimum procedure reduces to one which uses correlation or matched-filter techniques. The analysis applies not only to the specific problem described here, but to any situation in which the patterns to be classified may be represented by binary vectors provided the appropriate assumptions are satisfied. An important example is the so-called property-list description of a pattern.
© 1965 Optical Society of America
Full Article | PDF ArticleMore Like This
Thomas J. Harley and Laveen N. Kanal
Appl. Opt. 4(10) 1351-1352 (1965)
George Eichmann and Sanghamitra Basu
Appl. Opt. 26(10) 1859-1865 (1987)
Andrew Guttman
Appl. Opt. 4(1) 91-98 (1965)