Page 404 - Introduction to Information Optics

7.5. Pattern Recognition Using Composite Filtering 389

Fig. 7.30. (a) Input scene, (b) Output correlation distribution.

where

DR = ^l^^ (7.33)
'antitarget — max

is known as the target discrimination ratio, / target -min is the minimum correla-
tion peak intensity of the target set, and /antitarget -max is the maximum
correlation peak intensity of the antitarget set. The reason for using the
combined criteria is that, although the correlation peak intensity (CPI) is
primarily used to evaluate the capacity of the composite filters, we also have
to consider the discrimination ability of these filters; that is, the DR should also
be used as a criterion to evaluate the capacity of the filters. Thus, we see that
these two criteria will be simultaneously used for the evaluation of performance
capacity.
To investigate performance capacity, a set of 32 x 32, 64 x 64, and 128 x 128
pixel-element SABCFs are synthesized using the same target and antitarget
images. This set of SABCFs is constructed with and without the antitarget set,
respectively. By implementing this set of SABCFs respective to the JTC,
normalized correlation peak intensities (CPIs) as a function of number of
training images have been plotted. We have seen that the CPIs montonically