
7.3
Least Squares Filtering
281
for some w
0
. Of course, w
0
should be chosen so as to exclude the zeros of
H(u, v) from the neighborhood. An example of restoration done in this way
is shown in Figs. 4a-4e. The neighborhood used in Fig. 4e is small enough to
avoid the zeros of H(u, v), so that the restoration is fairly good. When a larger
neighborhood is used, the results become very poor, as shown in Fig. 4f.
7.3
LEAST SQUARES
FILTERING
One way to avoid arbitrariness in inverse filtering is to find a restoration
f(x,y) of the picture f(x, y) that minimizes some measure of the difference
between f(x, y) and f(x, y). If the restored pictures ar ...