(173)
deals with objects of low contrast is also well known. Carman and Carruthers'
have shown that the distribution of contrast in the object space is heavily
skewed toward the low-contrast detail; and so the lower the element (A c) (A x)
is positioned in the contrast domain, the more symbols it contains. The
Carman-Can uthers results are shown in Figure 4. Because of this, any lowering
of the contrast threshold for any size of symbol becomes most important in
terms of the number of symbols that will then be located above the threshold.
We now examine axial contrast reduction characteristics for the six-inch
Metrogon used in conjunction with Super-XX at each of three focal settings.
Fig. 4. Distribution of Contrasts in Aerial Scenes.
These data are then expressed in terms of threshold characteristics of the
Metrogon-Super-XX system. Above these thresholds a symbol will be detected;
below, the symbol will remain undetected. These characteristics are shown in
Figure 5. It is clear from this plot that focusing for maximum resolution pro
vides for the detection of the smallest symbols that the system can record. It
is also clear, however, that certain larger symbols are more apt to be recorded
at other focal settings since at these settings lower thresholds are indicated for
these particular sizes.
We rarely achieve in airborne use resolution values that correspond to
laboratory performance. Thus we are dealing with larger symbols than those
examined at the resolution limit in the laboratory. It is also true that in
practice we are not interested in just one size of symbol. Both these facts
suggest that, until the distribution of significant symbol sizes be known ’), the
best focal setting we can select is that which provides a weighted minimum
threshold over the entire range of sizes explored by the system. The weighted
area of the object space above this threshold, i.e., the size of the object domain
available to exploration by the system, may then be regarded as a measure of
the performance.
This weighting of the object space area may be done directly from the
i) There has been evidence presented (3) from which one may infer that the distribution in
the object space of significant symbol sizes will reveal no requirement for weighting any one
size over another.