60
60
abi GEHEN dU e ER NS 0 7 PR TR MUR eL STEIN DER AERE €
19.
nents of the photogrammetric system and the inclination of the incident
ray. These sizes were approximately 1 to 2 times the widths of the
spread functions of the systems. Detectability of objects is clearly
a function of target contrast, imaging quality of the photogrammetric
system and optical magnification of observation. Welch and Halliday's
figures were derived at a variety of optical magnifications, and
therefore the size of 10-30um represents a good overall estimate of
object detectability for the image qualities involved.
They extended their work on photogrammetric targets to investigate
the measurability of objects. i.e. the threshold size of the target
for reliable measurement. In this case the sizes of the targets varied
between 20um and 60um also depending on system characteristics. A
factor of approximately 2 between sizes of detectability and recog-
nition appears to be relevant in this study. This same ratio can be
found in the comparison of detectability and recognition given in the
previous section, while the relationship between spread function
width and recognition is also a constant factor of approximately 2 to
4. Insufficient data is available to investigate the FL for these
targets.
4.3 Subjective Assessments of Image Quality for Purposes of
Interpretation
Hufnagel (1965) and Scott (1968) performed experiments to deter-
mine the image quality parameters of a photographic image which are
significant for the purposes of interpretation. A series of photo-
graphs of varying image qualities were presented to many observers for
ranking in order of quality. The MTF's of the photographs were
Gaussian in shape for one set, and Gaussian with a "hump" superimposed
for a second set. Varying degrees of graininess were superimposed on
different photographs. During the ranking of the photographs, the
observers were asked to compare grainless and grainy photographs with
different MTF's.
In assessing the results Hugnagel found that grainless scenes
were ranked of equal quality when their Transfer Functions were
approximately equal at high frequencies irrespective of their shapes
at low frequencies. Further, he found that the range of modulations
for which frequencies were significant was between 0.1 and 0.4.
However, if grain existed the range of significant modulations in-
creased until for extreme cases of grain the ranking became indepen-
dent of MTF and dependent on contrast of the object and other factors
such as non-linearity of the tonal response. Significantly (Frieser
and Biedermann, 1963) adopted the frequency for which the MTF curve
equalled 0.5,which agrees with the region of significance of Hugnagel
if granularity exists.
The conclusions of Hufnagel give some confirmation to the use of
the frequency limit as a parameter for expressing pointing precision,
P |
Rm