Abstract
In a recent paper1 we introduced a novel procedure for the generation of spatial filters optimized with respect to the entropy of the correlation plane intensity distribution. Compared with the conventional matched filter, the entropy optimized filter indicated superior pattern recognition performance even in the presence of substantial noise in an apparent contradiction with theoretical expectations. Employing procedures of communication theory this apparent contradiction is traced back to the idealized definitions related to the matched filter. If practical limitations are taken into account the entropy optimized filter and other synthetically generated filters can be predicted to perform better than the matched filter.
© 1989 Optical Society of America
PDF ArticleMore Like This
Uri Mahlab, Michael Fleisher, and Joseph Shamir
TuI17 Optical Computing (IP) 1989
John H. Hong and Pochi Yeh
TuI22 Optical Computing (IP) 1989
Liang Minhua, Wu Shudong, Liu Liren, and Wang Zhijiang
TuI26 Optical Computing (IP) 1989