ENGINEERING REPORT NO. 8976 PSYCHOPHYSICAL GEMS STUDY FINAL REPORT

• Document Type: CREST
• Collection: NGA Records (Formerly NIMA)
• Document Number (FOIA) /ESDN (CREST): CIA-RDP78B04747A000700010023-9
• Release Decision: RIPPUB
• Original Classification: S
• Document Page Count: 55
• Document Creation Date: December 28, 2016
• Document Release Date: March 29, 2002
• Sequence Number: 23
• Publication Date: October 16, 1967
• Content Type: REPORT

Approved For Release 2002/06/17 : CIA-RDP78B04747A000700010023-9
25X1A
Declass Review by
NIMA/DOD
Approved For Rele =QII2/1(17-CIA-RDP78B04747A000700010023-9
L, +'' [ V S H - M x - o o-7
~ a.
 40 '
25X1A
Approved For Release 2002/06/17 : CIA-RDP78B04747A000700010023-9
ENGINEERING REPORT NO.
PSYCHOPHYSICAL GEMS STUDY
29(1 A
SPO 27203
DATE: October 16,__1967
Approved For Release 2002106/17 : CIA-RDP78BO4747A000700010023-9
vsNk - mX - ??9 - l
 Section
Approved For Release 200 /
-RDP78B04747A000700010023-9
jQQ
TABLE OF CONTENTS
Title
Page
ABSTRACT ii
PSYCHOPHYSICAL STUDY INTRODUCTION
1.1 General
1.2 Parameters of GEMS
1.3 GEMS Viewing Factor
DISCUSSION OF GEMS STUDY RESULTS
2.1 Statement of Study objectives
2.2 Results of Experiment I
2.3 Results of Experiment II
2.4 Results of Experiment III
2.5 Realism of Simulations
STUDY CONCLUSIONS
3.1 Study Implications
3.2 GEMS - An Evaluation Tool
APPENDIX A - Psychophysical GEMS Study Report
APPENDIX B - Psychophysical GEMS Matrix Data
Approved For Release 2002/06/1k' : CIA-RDP78B04747A000700010023-9
11
 Approved For Release 200011aA-RDP78B04747A000700010023-9
INO
ABSTRACT
A psychophysical study was conducted to determine the usefulness of
GEMS as a subjective image quality evaluation tool. The study results definitely
established the fact that GEMS can be employed to obtain realistic estimates of
system performance. The study experiments provided valuable information con-
cerning the matrix increment spacings for modulation transfer function and
ground exposure. The data also indicates that the GEMS parameters are unique.
ii
Approved For Release 2002/06/17 :. CIA-RDP78BO4747A000700010023-9
et T~
 '8 3 R'd
Approved For Release 200 -RDP78BO4747A000700010023-9
25X1A
A6
25X1A
PSYCHOPHYSICAL STUDY INTRODUCTION
1.1 GENERAL
A program task of utmost importance is the Psychophysical GEMS study.
The prime objective of the study is to determine the usefulness of GEMS as a
means of subjectively evaluating mission material. A study was performed with
three GEMS matrices at the customer's facility by
The experiments conducted by
25X1A
were 25X1A
designed to gain an insight on (1) how fine an increment spacing of modulation
transfer function (MTF) and ground exposure could be distinguished by photo-
interpreters, (2) the uniqueness of each parameter, (3) the accuracy and
repeatability of judgments, and (4) the ability of interpreters to rank mission
material with GEMS.
study report is included in this report
as Appendix A. The study results have been reviewed by
and an 25X1 A
interpretation of these results is discussed in the following sections of this
report. Before proceeding with a discussion of the tests and the results, a
basic understanding of the stimuli material is essential.
1.2 PARAMETERS OF GEMS
Matrix arrays of MTF and exposure were generated for three different
type scenes. The parameters of these matrices were defined in conjunction with
the customer and also based on the findings of the System Parameters study.
The ten levels of simulated MTF were inclusive of a 6.3 to 16.2 foot ground
resolution range at 10% increments of resolution. It was the intent of the
study to determine the increment spacing for MTF above and below the present
best performance level of the system.
Approved For Release 2002/06/17 : CIA-RDP78BO4747A000700010023-9
 Approved For Release 2002/064t
RP781304747A000700010023-9
The seven increments of ground exposure were simulated to represent
a total exposure shift of 0.38 density units. This simulated exposure range
corresponds to approximately a + 0.20 density shift on either side of the
average ground exposure value determined in the System Parameters study. The
increment spacing was varied from 0.02 to 0.15 density units. The exact values
of each increment can be obtained from Appendix B, the Psychophysical GEMS
Matrix data report.
The scene content of the GEMS consisted of a harbor, an airfield,
and a city area. The scale of the imagery was a fixed parameter for the
experiment.
1.3 GEMS VIEWING FACTOR
Due to the problems of controlling MTF above 100 cycles per milli-
meter in the simulation process and obtaining original material with the
proper photographic parameters, the psychophysical GEMS were generated at three
times the scene scale of the appropriate mission material. The desired scene
scale factor was simulated by viewing the mission material at three times the
viewing magnification of the GEMS. To account for any apparent differences in
image grain size, the GEMS were simulated on a film type that visually appeared
to yield the same grain structure as type 3404 film at a factor of three less
viewing magnification. The spatial frequency extent of the simulated MTFs
also were generated to account for this difference in viewing magnification.
Approved For Release 2002/06/17 2 CIA l ',r$ 47A000700010023-9
 2.1 STATEMENT OF STUDY OBJECTIVES
The psychophysical tests were divided into three experiments. The
first experiment involved only the use of GEMS. Experiment I was concerned
with psychologically scaling the parameter of MTF. The purpose of the experi-
ment was to obtain the necessary information required to specify the equal
visual increment spacing for this parameter.
DISCUSSION OF GEMS STUDY RESULTS
Experiment II was designed to determine the psychological scaling of
the ground exposure parameter and to test how accurately and reliably subjects
could judge the parameters of MTF and exposure. The tests also were designed
to determine if the parameters of the matrix array were unique. In other words,
that there were no two elements of a matrix array that produced the same equiva-
lent impression of MTF and exposure.
Experiment III was conducted to determine how reliably subjects could
judge the MTF and exposure levels of mission material with GEMS. The accuracy
of these judgments could not be determined since no objective measurements were
made of the imagery used in the tests.
Approved For Release 20 O(/1i i&IA-RDP78B04747A000700010023-9
The details pertaining to methods and procedures of each experiment
have been thoroughly covered in the
Ireport. The text
no*
25X1A of the
report will be confined to a discussion and an interpre-
tation of the study results with a few minor exceptions.
25X1A
Approved For Release 2002/06/17 : IA-RDP78B04747A00.0700010023-9
 Approved For Release 20021
2.2 RESULTS OF EXPERIMENT I
2.2.1 MTF Spacing
p1AIRDP78BO4747A000700010023-9
The results of Experiment I indicate that equal visual incre-
ments changes of MTF are obtained when the MTF parameter is equally spaced
logarithmically. The experiment also indicated that a 677 discrimination of
MTF is obtainable with an increment spacing corresponding to 1.4 log steps. A
1.4 log step is equivalent to a 14% change in ground resolution.
2.2.2 Definition of Resolution. Terms
i
To avoid any misunderstanding between the use of the terms
MTF and ground resolution, it will be necessary to define the meaning of these
terms in relationship to how they apply to the study. MTF spatial frequency
is a parameter which is dependent upon the viewing magnification factor. Thus,
when an observer views 50 lines per millimeter imagery with a 10 power micro-
scope, the spatial frequency extent of the imagery observed by the eye is only
5 lines per millimeter. In simplified terms, this means that the film imagery
MTF is just translated from a 50 to a 5 line per millimeter extent. Visually,
the spread function size of the imagery is just enlarged by the magnification
factor of the optics, provided the optics of the microscope do not introduce
additional degradations.
However, when speaking of the ground resolution (in feet) of
the imagery, the ground resolution value remains the same regardless of the
viewing magnification factor and provided no degradations are introduced by
the optics. Since the simulation of MTF was made to be dependent upon viewing
magnification, it was perceived that it would be best to express the MTF results
in, terms of ground resolution to avoid any confusion that might result due to
the use of the various magnification factors. When the 10% increment values
Approved For Release 2002/06/1 : CIA-RDp$F~04747A000700010023-9
 Approved For Release 200 Tt1 cl4-RDP78B04747A000700010023-9
were established for the GEMS matrix MTF spacing, the increments were expressed
in terms of limiting spatial frequency resolutions. These limiting resolution
values (in lines per millimeter) can be directly converted into ground resolu-
tion values (feet) with the use of the imagery Scale factor. Therefore, when
discussing a 14% change in ground resolution, the equivalent change in limiting
spatial frequency resolution is also 14%.
The increment spacing of MTF was based on the limiting resolu-
tion values to avoid the ambiguity of specifying percentage changes in MTF
curve position at some arbitrary modulation level.
2.3 RESULTS OF EXPERIMENT II
2.3.1 MTF and Exposure Scaling
The findings of Experiment II verified that the MTF increments
of GEMS should be established at equal logarithmic steps. To obtain 67% dis-
crimination, the MTF steps should correspond to 12.5% increments of ground
resolution. It may be recalled that the results of Experiment I indicated 14%
increments. Since the population of subjects is greater for Experiment II, it
seems reasonable to accept the 12.5% number as a measure of the MTF increment
spacing.
Experiment II also implies that 67% discrimination of exposure
shift is obtained with an increment spacing of 0.06 density units.
2.3.2 Uniqueness of Parameters
One objective of the Psychophysical GEMS study was to deter-
mine that the elements of the GEMS matrix were unique. In order to obtain a
valid estimate of system performance with a GEMS matrix array, it is important
that no two elements in an array will be judged as being equivalent in MTF and
considered surprisingly low, and it exhibits the measurement accuracy and
simulation control for the entire GEMS simulation process. As much as a + 15%
variation in MTF existed among the original three scenes, and the variation
over the format of any one scene was just as great. An attempt was made to
minimize the MTF variations of the original scenes when generating the master
transparencies. If one takes into account that additional errors are intro-
duced by two independent MTF evaluation steps for each scene and that there is
a + 1% variability in the MTF simulation process among scenes, a total ground
resolution simulation variability error of + 6% could realistically exist.
The tests indicate that the exposure level of Scene A, the
harbor area, was approximately 0.11 density units denser than Scene C. An
explanation for this discrepancy can be given. The average density value for
each GEMS scene at a common exposure level was established to be the same when
evaluated. In the case of the harbor scene, the average density value was
obtained from density readings of both the image structure and the water. Since
the water is represented in the GEMS by low densities, it follows that the image
6
Approved For Release 2002/06/17 : Cl RPP7r8B 4747A000700010023-9
 Approved For Release 2002/06614 i4. P78BO4747A000700010023-9
structure of the harbor scene must be higher in density in order to achieve an.
Average scene density equivalent to that obtained from a transparency that only
consisted of image structure.
Actually, the average densities for the three scenes should
have been determined from imagery common to all scenes. During the tests, the
subjects matched the exposure levels by examining areas in the scenes with
common imagery and not by obtaining an integrated effect of the whole scene.
When considering that the resolution of the imagery is exposure dependent, the
procedure employed by the subjects to match exposure is most valid. Re-evalua-
tion of only the image structure areas of Scenes A and C verified the fact that
the average image structure density of Scene A was greater than Scene C. The
re-evaluation established the average density of Scene A to be denser by 0.14
density units. The reported value of 0.11 density units is in agreement.
2.3.4 Parameter Increment Spacin&
To summarize, the findings of the study indicate that a 67%
discrimination of MTF and exposure can be obtained if the MTF increments are
12.5% ground resolution differences and the exposure increments are 0.06 density
unit differences. The discrimination level of each parameter is not the only
criterion to be examined in establishing increment spacing. The discrimination
levels sometimes can be too coarse or too fine to serve as an efficient and/or
effective evaluation tool.
It is believed that the exposure increments, if established
at 0.06 density differences, would be too fine an increment spacing. The test
results imply that, for the 0.38 density exposure range simulated, little or
no change in resolution was observed due to the total exposure shift. This
test result was substantiated by visual examination of a medium contrast AFBT
which was employed in the GEMS' control and evaluation target array. Inspection
Approved For Release 2002/06/17 : CIA-RDP78BO4747A000700010023-9
7
 Approved For Release 2002/06/17e i01A-RDP78BO4747A000700010023-9 .
of the AFBT at both ends of the exposure range indicated that the resolution
did not change by a full AFBT element. Simultaneous viewing of the same two
AFBT elements, at each'end of the exposure series, only yielded the impression
that the over-exposed element was slightly more resolvable than the under-
exposed element.
If a matrix array with an exposure parameter is to serve as a
useful evaluation tool, the increment spacing of exposure should be established
on the basis of the degree of exposure shift that produces a 12.57 change in
limiting ground resolution. When specifying an exposure increment correspond-
ing to a 12.5% difference in ground resolution, it should be noted. that this
increment spacing is-not acquired by introducing a shift in the MTF; but that
it solely results from changes in the contrast of the imagery and the film
granularity at the various exposure levels. Since the change in limiting
ground resolution is not a linear function of exposure, the density increment
shifts of exposure will not be equally spaced. For the camera system and film
type involved, 4 to 6 exposure levels would be required for the GEMS matrix.
It also is suggested that the MTF parameter be established at
a 12.5% ground resolution increment spacing. For measures of mission material
quality, as fine an increment spacing that can be visually detected, is desirable.
When making comparisons with the use of a split field microscope, it is possible
that the MTF discrimination level at this increment spacing could be greater
than 67%.
2.4 RESULTS OF EXPERIMENT III
A GEMS matrix array was employed with a number of mission material
scenes to determine how well different subjects agreed on their judgments of
the. MTF and exposure levels of the mission photographs. Standard deviations
Approved For Release 2002/06/17 : CIA-
17 047 7A000700010023-9
.. 1 F
Plip,
~,,a d
 Approved For Release 2002/06/t 781304747A000700010023-9
were computed from the subjects' responses for.each mission scene. The mean
pf the standard deviations for the MTF levels of the mission photographs was
computed to be 19% as opposed to a standard deviation of 12.5% when matching
one GEMS against another in Experiment II. The difference between the two
standard deviation values was not statistically significant.
The computed increase in the MTF standard deviation may be related
to the general comments stated by the subjects during the Experiment III test.
When matching a photograph with a GEMS, some subjects found a match in limit-
ing resolution detail, but observed that the edge sharpness between the two
.photographs did not quite agree. This comment signifies that the shapes of
the two photographic image spread functions, as observed by the eye, were not
identical. If this condition prevailed and some subjects based their judgments
on limiting resolution and others based their judgments on edge sharpness, it
would be anticipated that the mean standard deviation value would increase.
The observed difference between limiting resolution and edge sharpness is
either associated with the simulated MTF curve shape and/or with the differences
in the MTF of the two viewer optical systems.
The mean of the standard deviations for the exposure levels of the
mission photographs was 0.065 density units. The standard deviation of exposure
with the GEMS was computed to be 0.06 density units. Results of this nature
represent excellent agreement.
It can be concluded that the agreement among subjects as to their
judgments of the MTF and exposure levels of the mission photographs was approxi-
mately as good as with the GEMS. Nothing can be stated about the accuracy of
their judgments since no objective measures exist for the mission material
imagery used in the study.
Approved For Release 2002/0@/17 =w7B04747A000700010023-9
 Approved For Release 2002/061
25X1A
2.5 REALISM OF SIMULATIONS
Outside of the comment discussed in Paragraph 2.4 about the observed.
differences in limiting ground resolution and edge sharpness, the subjects
thought that the simulations were quite realistic of the appearance of mission
material.
The subjects suggested that the scale of. the buildings did not seem
to match the mission material even though the cars and streets looked identical
in size. It is believed that the noted difference in building scale size was
the result of observing the imagery with two different microscope fields of
view. S inAop &eg r F*F&sviz2661/0 /1f : 9IAt ' 7 ~1 '`A~tf? OOt1 23'-old of
view was such that a few entire buildings were o d by the eve- where.- at
 Approved For Release 20026,17 ~ c RDP78BO4747A000700010023-9
SECTION III
STUDY CONCLUSIONS
3.1 STUDY IMPLICATIONS
3.1.1 Parameter Scaling
The study results indicate that equal, visually discriminable
steps of MTF and exposure are obtained for the ranges of the parameters tested
if the increments of ground resolution are spaced at equal logarithmic steps
and the increments of exposure shift are spaced at equal density steps. A 67%
accuracy of discrimination is obtained for both parameters if the MTF steps
are established at 12.5% resolution intervals and the exposure steps are
established at 0.06 density intervals.
The results imply that an exposure shift interval of 0.06
density units is too fine for image quality evaluation purposes since no changes
in resolution due to exposure are discernable for the entire simulated exposure
range of the study GEMS. For the GEM matrix array, it is suggested that the
exposure increments be established to correspond to exposure shifts that pro-
duce 12.5% changes in limiting ground resolution.
An MTF interval of 12.5% ground resolution is as fine an
increment spacing as can be discriminated with reasonable accuracy. It is
suggested that this MTF interval be established as the increment spacing for
the GEMS matrix array. With the use of a split field viewer for making com-
parisons, the accuracy of MTF discrimination may be improved.
3.1.2 Other Study Findings
The experiment concerning the ranking of mission material
with a GEMS matrix indicated that the subjects' judgements of the MTF and
Approved For Release 2002/06/17 :1CIA-RE[3041"'1A000700010023-9
 Approved For Release 2002/ gh4-RDP78B04747A000700010023-9
exposure levels of the mission photographs were just as reliable as their judg-
ments when ranking one GEMS against another. In other words, the mean standard
deviations of their judgments for each parameter were statistically the same
in both experiments. The accuracy of their judgments could not be determined
since no objective measures were made of the mission photographs used in the
study.
A very interesting and important finding of the study was
that scene content seemed to have no effect on the subjects' judgments of MTF
or exposure. This finding strongly suggests that the parameter of scene
content can be confined to few scenes of various type imagery.
3.2 GEMS - AN EVALUATION TOOL
All indications of the Psychophysical GEMS study point to the fact
that GEMS can be employed in the evaluation of mission material. The results
of the study show that photo-interpreters can distinguish and rank independ-
ently the parameters of exposure and MTF. Their ability to discriminate levels
of exposure is better than is necessary. Although their discrimination of MTF
levels may be slightly less sensitive than desired, reasonable estimates of
MTF levels are obtainable with GEMS.
The major attributes of GEMS are that they can supply estimates of
system performance quite rapidly and that these estimates are not dependent
upon either specific type targets or complicated instrumentation that requires
lengthy operator training time. It probably would require a minimum of 4 to 8
.hours to objectively evaluate, without an on-line computer to the instrumenta-
tion, the parameters of a single scene area that could be estimated with a
GEMS matrix in a matter of a few minutes.
12
Approved For Release 2002/06/17 : CIA-RDP78BO4747A000700010023-9
 `4 n
Approved For Release 2002106/{7 A Fk 78B04747A000700010023-9
A point worthy of serious consideration is that the GEMS estimates
,of MT.F are almost as accurate as the MTF objective determination of mission
material, because of the variable factors associated with the typical targets
used in the measurement process. Under certain conditions, where the quality
of imagery is varying over the format and no objective measurement targets
exist in the areas of interest, the MTF estimates obtained with GEMS can be
much more accurate than any inferred evaluations. Image quality variations
occur quite frequently due to variable haze conditions or variable cloud
coverage exposure conditions.
In summary, the psychophysical study demonstrated that GEMS can
provide a useful service as a rapid evaluation tool.
Approved For Release 2002/06/17 : QI 1-R[ 78BO47n4,7A000700010023-9
 Approved For Release 2002/4FiIZ ,i#IA-RDP78B04747A000700010023-9
25X1A
Psychophysical GEMS Study Report
Approved For Release 2002/0(9i7T; C1Ay1 DP78B04747A000700010023-9
L
 Approved For Release 2002/06/17 : CIA-RDP78B04747A000700010023-9
r.tiF4 LLh~
THE JUDGMENT OF GROUND RESOLUTION
AND EXPOSURE SHIFT OF AERIAL PHOTOGRAPHS
25X1A
25X1A
25X1A
October 2, 1967
25X1A
E'7 ~ .~ =1 a?11
Approved For Release 2002/r17 `: C.iii-RDP78B04747A000700010023-9
l }. P
 ~g fF '~ j fy ev u
Approved For Release 20024I2 , ., (k- DP78BO4747A000700010023-9
The assistance given byl I 25X1A
is very much appreciated. He, with the aid of
T_ _ personnel, supervised the experimental
sessions. The author wishes also to thank the
staff members of T__ _, who patiently spent
hours of their time serving as subjects.
ii
Approved For Release 2002/0 ' X4` 9P78BO4747A000700010023-9
 Approved For Release 2002/06JFi+ 78B04747A000700010023-9
J -H
Page
ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . . ii
LIST OF TABLES AND FIGURES . . . . . . . . . . . . . . iv
STUDY OBJECTIVES . . . . . . . . . . . . . . . . . . . 1
EXPERIMENT I . . . . . . . . . . . . . . . . . . . . . 2
Purpose . . . . . . . . . . . . . . . . . . . . . . 2
Method . . . . . . . . . . . . . . . . . . . . . 2
GEMS . . . . . . . . . . . . . . . . . . 2
Subjects and procedure . . . . . . . . . . . . . 2
Results and Discussion . . . . . . . . . . . . . . 4
EXPERIMENT II. . . . .. . . . . . . . . . . . . . . . 7
Purpose . . . . . . . . . . . . . . . . . . . . . 7
Method . . . . . . . . . 7
. . . . . . . . . . . .
GEMS . . . . . . . . . . . . . . . . 7
Experimental design . . . . . . . . . . . . . . . 9
Subjects and procedure . . . . . . . . . . . . . 9
Results and Discussion . . . . . . . . . . . . . . 9
Scaling . . . . . . . . . . . . . . . . . 9
Accuracy of judgment . . . . . . . . . . . . . 13
Reliability of judgment. . . . . . . . . . . . 20
EXPERIMENT III . . . . . . . . . . . . . . . . 23
Purpose . . . . . . . . . . . . . . . . . . . . . . 23
Method . . . . . . . . . . . . . . . . . . . . 23
Photographs . . . . . . . . . . . . . . . . . . . 23
Subjects and procedure . . . . . . . . . . . . . 23
Results and Discussion . . . . . . . . . . . . . . 23
IMPLICATIONS . . . . . . . . . . . . . . . . . . . . 25
iii
Approved For Release 2002/06/1'4 Rgp78B04747A000700010023-9
 Approved For Release 2002I067 P78B04747A000700010023-9
LIST OF TABLES AND FIGURES
Page
1 THE GROUND RESOLUTION AND EXPOSURE SHIFT OF
THE GEMS . . . . . . . . . . . . . . . . . . . . 3
2 EXPERIMENTAL DESIGN . . . . . . .. . . . . . 10
3 ANALYSIS OF VARIANCE OF ERRORS OF EXPOSURE-
SHIFT JUDGMENTS . . . . . . . . . . . . . . . . 14
4 ANALYSIS OF VARIANCE OF ERRORS OF GROUND-
RESOLUTION JUDGMENTS . . . . . . . . . . . . . . 17
5 MEANS OF THE STANDARD DEVIATIONS. . . . . . . . 22
?6 MEANS AND STANDARD DEVIATIONS OF MISSION AND
PHOTOGRAPH JUDGMENTS . . . . . . . . . . . . . . 24
Figure
1 Proportion of judgments of "better resolution"
as a function of ground resolution. . . . . . 5
2 The relation between ground resolution and
judgments in Z units. All three series from
Figure 1 combined . . . . . . . . . . . . . . . 8
3 The relation between ground resolution and
visual scale values . . . . . . . . . . . . . 11
4 The relation between exposure-shift and
visual scale values . . . . . . . . . . . . . . 12
5 Mean error of exposure-shift judgments as a
function of exposure shift and scenes . . . . . 12
6 Mean error of exposure-shift judgments as a
function of ground resolution and scenes. . . . 16
7 Mean error of ground-resolution judgments as
a function of resolution and scenes . . . . . . 18
8 Mean error of ground-resolution judgments as
a function of exposure shift and scenes . . . . 20
iv
Approved For Release 2002/7, CIA-E2DP78B04747A000700010023-9
r, ii
 Approved For Release 2002/06/17 : CIA-RDP78B04747A000700010023-9
STUDY OBJECTIVES
The goal of this study was to determine how well humans
can judge the quality of aerial photographs by having them
compare one set of photographs with a standard set. -Al-
though there is a large number of determinants of photo-
graphic quality, this study was concerned with only two of
them--ground resolution and exposure shift.
To achieve this goal, aerial photographs were produced
with ground-resolution and exposure-shift values. These
photographs (hereafter referred to as GEMS) and mission
(operational) photographs were used in a series of three ex-
periments to achieve the following objectives:
1. To scale psychologically exposure shift and
ground resolution. The purpose of this
scaling was to determine the relation be-
'tween physical measures of exposure shift
and ground resolution on the one hand, and
psychological values on the other. Psy-
chological scaling provides the information
necessary to space the GEMS at equal visual
increments.
2. To determine how accurately and reliably
trained subjects can judge the exposure
shift and ground resolution of photographs.
3. To determine whether or not the judgment of
ground resolution and exposure shift is vi-
sually independent. That is, to determine
whether the exposure-shift value of a photo-
graph affects the accuracy of ground-
resolution judgments, and to determine
whether ground resolution of a photograph
affects exposure-shift judgments.
4. To determine how reliably subjects can judge
the exposure shift and ground resolution of
mission (operational) photographs. Accuracy
of judgment could not be assessed because no
exact specifications of ground resolution
and exposure shift were available for the
aerial photographs.
Approved For Release 2002/06/ ` k C~,r i D
 Approved For Release 2002/P6
EXPERT,MENT I
Purpose
The primary purpose of Experiment I was to determine
whether the difference among the ground resolutions of
the GEMS were sufficiently di,scr.mi,nable to be used i,n
Experiments IT and III. The secondary purposes of Experi-
ment I were to psychologically scale ground resolution and
to establish a "just noticeable difference'l for ground re-
solution, that is, a reliably discriminated difference in
ground resolution.
GEMS. Two sets of GEMS were used in this experiment,
a set of a harbor scene (Scene A) and a set of an airfield
(Scene B). All of the GEMS were negative transparencies.
Table I (Page 3) shows the exposure shift and ground resolu-
tion of the 70 Scene-B GEMS and 21 Scene-A GEMS. Each cell
in the table represents one Scene-B GEM with a particular
combination of ground resolution and exposure shift. (The
cells marked with Xs refer to the GEMS used i.n Experiment
II.)
Each Scene-A GEM was a combination of one of the seven
exposure-shift values with one of three ground resolutions--
13.1, 10.71. and 7.8 feet. The density value of .00 refers
to an underexposed-(light) GEM, and the density value of .38
refers to an overexposed (dark) GEM.
Subjects and procedures. The six subjects who parti,ci
pated in this study were experienced in viewing aerial
photographs. Their experience ranged between one and ten
years with a median of 3.5 years. Each of the 21 GEMS from
Scene A served asa standard for judging a series of five
GEMS from Scene B. The selection of the Scene-A GEMS to
serve as standards was arbitrary. In each series, one of
the GEMS was equal, two were worse, and the other two were
better than the ground resolution of the Scene-A GEM. For
example, one of the standards from Scene A had a resolution
of 13.1 feet at a density of..00; the resolutions of the
Scene-B GEMS were 16.2, 14.6, 13.1, 11.8, and 10.7 feet.
Each subject, then, made a total of 105 judgments, five
judgments for each of 21 series.
The Scene-B and Scene-A GEMS were assigned a random
number. Each subject was given a response sheet with the
order in which he was to judge the GEMS. His task was to
select the standard from Scene A and the comparison GEM
Approved For Release 2002/06/17 : dA-RDP78B04747A000700010023-9
 Approved For Release 2002/06/17 : CIA-RDP78BO4747A000700010023-9
TABLE 1
THE GROUND RESOLUTION AND EXPOSURE SHIFT OF THE GEMS
GROUND RESOLUTION (FEET)
16.2
14.6
13.1
11.8
10.7
9.6
8.6
7.8
7.0
6.3
.00
02
u
X
X
X
X
em
, .
=z .05
X
X
X
X
~~ .08
X
X
X
X
13
z
X
X
X
X
.
o
w w
w0 .23
X
X
X
X
.38
Approved For Release 2002/06/17 : CIA-RDP78BO4747A000700010023-9
 Approved For Release 2002/06 T', ,ICI r-A P78B04747A000700010023-9
from Scene B, and place each on the two stages of a
High-Power Stereo-Viewer. Viewing one GEM through one
ocular of the microscope, and the other GEM through the
other ocular, he alternated between two eyes. The subject
judged whether the Scene-B GEM was better or worse in re-
solution than the Scene-A GEM. If he judged the Scene-B
GEM better, he circled the number of the GEM on his re-
sponse sheet. If its resolution was worse than the Scene-
A GEM, he did not circle the number.
The order of the GEMS in a series, the three series at
each density level, and the seven sets of three series (one
set at each density), was randomized separately for each
subject.
Each subject worked at his own rate, and rested when-
ever he felt tired.
Results and Discussion
Preliminary analyses showed the exposure shift of the
GEM had no effect on the judgment of resolution. Therefore,
the judgments were pooled over levels of exposure shift.
The pooling resulted in three series with standards of 13.1,
10.7, and 7.8-feet ground resolution.
Figure 1 (Page 5) shows the proportion of "better re-
solution" as a function of ground resolution for each of
the three series. The results showed that discrimination
(indicated by the slope of the function) was about equal for
the three series. This result means that a logarithmic
spacing of ground resolutions leads to equally discriminable
steps throughout the series, at least when the ground re-
solution ranges from 16.2 to 6.3 feet as it did in this
experiment.
Inspection of the three series shows a peculiar result.
In all three series, the Scene-B GEM that was supposedly
equal in ground resolution to the standard was judged to be
better than the standard Scene-A GEM. The proportion of
judgments "better resolution" for that Scene-B GEM in Series
1 through 3 was approximately .57, :71, and..71. If the two
GEMS were judged to be equal, the proportion should have
been close to .50 for each of the series.
To determine if the discrepancy between the expected
proportion of .50 and the observed proportions was a reli-
able one, the proportion of judgments "better resolution"
were combined for the three Scene-B GEMS, the GEMS which
supposedly were equal in ground resolution to the Scene-A
GEMS. The mean of the judgments of "better resolution"
was .66. A critical ratio showed that there was a statis-
tically significant difference between the mean proportion
25X1A
Approved For Release 2002/06/17 : CLA-RDP78BO4747A000700010023-9
 Approved For Release 2002/0617-; CIA R P78BO4747A000700010023-9
(n
F- .70
W Z
O
CD H
pf-- .60
=D a
J
O
Li V)
ow .50
c,
Z
CD cc
F--I W
f-~- ..40
~F-
o w
a m
CD=
x .30
.00
16.2 14.6 13.1 11.8 10.7 9.6 8.6 7.8 7.0 6.3
(St) (St) (St)
GROUND RESOLUTION (FT.)
Fig. 1. Proportion of judgments of "better
resolution" as a function of ground resolu-
tion (St = standard Scene-A GEM; Jindicates
adjusted standards).
. 5
Approved For Release 2002/Q6/T7: CIA-RDP78B04747A000700010023-9
 Approved For Release 2002/06/1 7 :,c -.p 8BO4747A000700010023-9
of .66 and the expected proportion of .50 (CR = 12.53,
p < .01). This finding implies that the ground resolu-
tion of the Scene-B GEMS may have been better than that of
the Scene-A GEMS. Additional evidence for this conclusion.
will be presented in Experiment II.
Another peculiarity of the results shown in Figure 1
was the increase-in the proportion of judgments of "better
resolution," from Series 1 to 3. If, as the mean propor-
tions indicated, the standards were actually worse than
the ground resolutions specified on the abscissa in Figure
1, the results do seem reasonable and may be accounted for
in the following way: Assume the ground resolutions of the
standards in the three series were actually those values
(adjusted standards) indicated by the arrows on the abscissa
in Figure 1. (The values for the adjusted standards were
obtained from the results of Experiment II.) By assuming
this, the results shown in the figure may be due to the
commonly observed "central tendency" effect in psychophysics.
(The central tendency effect manifests itself as an under-
estimation of stimulus values at one end of a continuum and
overestimation at the other end.) In Series 1 and 2, the
adjusted standards were overestimated: The point of sub-
ject equalityl was slightly better than 13 feet in Series 1
(overestimated by 1.5 feet), and 10.7 feet in Series 2
(overestimated by less than one foot). Put in another way,
in relation to the adjusted standard, the ground resolu-
tions of the Scene-B GEMS were underestimated and the under-
estimation was larger for Series 1 than it was for Series 2.
In Series 3, the adjusted standard was underestimated; the
point of subject equality was 8.6 feet (underestimated less
than .5 feet). In contrast, then, the ground resolution of
the Scene-B GEMS in Series 3 were overestimated in relation
to the adjusted standard. This progressive change from
underestimation of the ground resolution of the Scene-B
GEMS in Series 1 to an overestimation of those in Series 3
resulted in an increase in the proportion of "better resolu-
tion" judgments from Series 1 to 3.
The above explanation, though speculative, is cer-
tainly reasonable. If the explanation is not acceptable,
the only other obvious explanation of the result is that
as the ground resolution improved there was a systematic
increase in the difference between the ground resolution of
the Scene-A and Scene-B GEMS.
To determine the relation between ground resolution and
discrimination, the judgments of the three series were
IThe point of subjective equality is the value on the ab-
scissa judged to be equal to the standard; i.e., the value
that elicits a proportion of .50.
Approved For Release 2002/06/17 : CIA-RDP78B04747A000700010023-9
 Approved For Release 2002/06/,j7c< f;l -P P78BO4747A000700010023-9
pooled. The proportion of judgments of "better resolution"
were combined and transformed to Z units (equal Z units
are assumed to be psychologically equal). Figure 2 (Page 8)
shows the relation between Z units and ground resolution.
(There is no data point, in Figure 2 for the best resolution
because the proportion for this resolution was 1.0 and the
Z value for 1.0 is infinite.) Figure 2 showed the relation
between ground resolution and psychologically equal inter-
vals is linear. This means that equal steps in log ground
resolution produced equal visual changes.
The figure also showed that a change of 1.4 log steps
in ground resolution was equal to one Z unit (the value of
1.4 is equal to one-half the range of minus one Z to plus
one Z value on the abscissa). A one Z value of 1.4 steps
in ground resolution means a difference this large was dis-
criminated correctly about 67% of the time.
The implications of results of this experiment are that
1) the ground resolution of GEMS should be equally spaced
logarithmically if they are to yield equal visible changes,
and 2) if 67% accuracy of discrimination is desired, the
change in the ground resolution of the GEMS should be about
1.4 log steps, where a step here was defined as a 10% de-
crease in ground resolution.
Purpose
The purposes of Experiment II were 1) to scale psycho-
logically exposure shift and ground resolution; 2) to deter-
mine the change in exposure shift and ground resolution re-
quired for the subjects to discriminate reliably; 3) to
determine the accuracy and reliability of judgment; and 4)
to determine whether judgments of exposure shift and ground
resolution are independent.
GEMS. GEMS of three scones were used in this study,
Scenes A and B (those described in Experiment I), and an
additional Scene C, an industrial site. Reference back to
Table 1 shows the exposure shift and ground resolution of the
Scenes-A and B GEMS. These are denoted by the cells con-
taining X. There were 20 GEMS of Scene A, 20 GEMS of Scene B,
and 70 GEMS of Scene C, one for each combination of exposure
shift and ground resolution shown in Table 1.
Approved For Release 2002/06/17 : CI4-RDP78BO4747A000700010023-9
 Approved For Release 2002/06/1
(; , , I 78B04747A000700010023-9
0.0
i
WORST
3 4 5
(ST) BEST
GROUND RESOLUTION
(EQUAL LOG STEPS)
Fig. 2. The relation between ground reso-
lution and judgments in Z units. All three
series from Figure 1 combined.
Approved For Release 2002/06/,t7x C>(F-RP78B04747A000700010023-9
 Approved For Release 2002/061 P78B04747A000700010023-9
Experimental design. Table 2 (Page 10) shows the ex-
perimental design. Each subject made two judgments of
Scene-A and Scene-B GEMS in relation to Scene-C GEMS. Five
subjects judged the Scene-A GEMS first and the Scene-B GEMS
second on Trial 1, and then judged the scenes in reverse
order on Trial 2. The other six subjects judged the Scene-B
GEMS first and the Scene-A GEMS second on Trial 1, and then
judged the scenes in reverse order on Trial 2.
The order in which each subject judged the GEMS within
a scene was randomized separately for each subject. The
time between the judgments of Scene-A and B GEMS was not con-
trolled--the subjects worked at their own pace. The minimum
time between trials was about one day.
Subjects and procedure. The 11 subjects who partici-
pated in this experiment were experienced at viewing aerial
photographs. Six of the 11 subjects had participated in
Experiment I. The range of experience was one to 11 years
with' a median of five years.
The 70 Scene-C GEMS were numbered and arranged on a
large light table in the form of a matrix as depicted in
Table 1. Each of the Scene-A and Scene-B GEMS was assigned
a random number.
Each subject was given a response sheet on which the
numbers of Scene-A and Scene-B GEMS appeared. Opposite each
number was a space provided for the number of the Scene-C
GEM which the subject felt best matched the Scene-A (or B)
GEM in both exposure shift and ground resolution. In other
words, the subject's task was to match the exposure shift and
ground resolution of each Scene-A and Scene-B GEM with that
of a Scene-C GEM.
The subject's task was as follows: He took the Scene-A
(or B) GEM, indicated on the response sheet, located its
exposure-shift value by matching it to one of the rows of the
matrix of Scene-C GEMS laid out on the light table next to
him. He then located the resolution of a Scene-C GEM that
best matched that of the Scene-A (or B) GEM he was judging.
Next, he placed the Scene-C GEM on one stage of a B&L High-
Power Stereo-Viewer and the Scene-A (or B) GEM on the other
stage of the microscope. The final judgment of which Scene-
C GEM best matched the A (or B) GEM in exposure shift and
ground resolution was made on the microscope. Of course,
the use of a stereo microscope required that the subject al-
ternate between eyes in making this judgment.
Results and Discussion
Scaling. The first analyses of the data from this ex-
periment were done to psychologically scale exposure shift
Approved For Release 2002/06/17 : CIA-RDP78B04747A000700010023-9
 TABLE 2
EXPERIMENTAL DESIGN
TRIAL
I
2
ORDER
I
2
3
4
5 SUBJECTS
A
B
B
A
6 SUBJECTS
B
A
A
B
Note: A&B Denote Scenes.
10
Approved For Release 2002/ 7, :C DP78B04747A000700010023-9
1, r7
 Approved For Release 2002/0 1?Z _' i,QI DP78BO4747A000700010023-9
w
? 1.5
0.93.1 10.7 8.6 7.0
GROUND RESOLUTION (FT.)
Fig. 3. The relation between ground-
resolution and visual scale values.
and ground resolution. The scaling of ground resolution
was accomplished by pooling the subjects' judgments of
ground resolution over trials, scenes and exposure, and
similarly, the scaling of exposure shift was accomplished
by pooling their judgments of exposure shift over trials,
scenes and ground resolutions.
The data were analyzed by a pair-comparison treat-
ment of categorical data.2
Figures 3 and 4 (Page 12) show the results of the
scaling for ground resolution and exposure shift.
Figure 3 shows a linear relation between scale values
and a log spacing of ground resolution (note that the
ground resolutions are spaced at equal log intervals).
Similarly Figure 4 shows a linear relation between scale
values and exposure shift in density units.3 (Note also
that density units are, in fact, on a log scale, so
that the spacing on the abscissa in Figure 4 is also a
log spacing.)
2For a description of this technique, see Guilford, J. P.,
Psychometric Methods. (2nd ed.) New York: McGraw-Hill,
1964. P. 242.
3A scale value for a density value of .23 was not computed
because the proportions for that value were close to 1.00.
Scale values should not be derived from proportions close
to . fd~p vet F%9 Release 2002/06/17 : Cl./\-RDP78B04747A000700010023-9
 Approved For Release 2002/06/1
.80
Fig. 4. The relation between exposure-shift and visual
scale values.
.00 .02 .05 .08
EXPOSURE SHIFT (DENSITY UNITS)
W CD
-. 03' I I - '
.00 .02 .05 .08 .13 .23
EXPOSURE SHIFT (DENSITY UNITS)
Fig. 5. Mean error of exposure-shift judgment'as a func-
tion of exposure-shift and scene. (88 judgments for each
data point.)
12
Approved For Release 2002/06/17 : CIA-RDP78BO4747A000700010023-9
 Approved For Release 2002/06/, G781304747A000700010023-9
1L
The implications of these findings are that the in-
crement in ground resolution of a GEM set should be set
at equal logarithmic values and that the increment of
exposure-shift values should be set at equal density
values.
Accuracy of judgment. An exposure-shift and ground-
resolution error was computed for each of the Scene-A and
Scene-B GEMS for each subject. The exposure-shift error
score was the density value of the Scene-C GEMS judged
equal to the Scene-A (or B) GEM, minus the density value
of the Scene-A (or B) GEM. Therefore, a positive error
means the Scene-A (or B) GEM was judged to be more dense
(overexposed), and a negative error means it was judged
to be less dense (underexposed) than the Scene-C GEM. The
ground-resolution error score was obtained by first assign-
ing the numbers 1-10 to the worst (16.2 feet) through the
best (6.3 feet) ground resolution of the Scene-C GEMS.
The numbers 8, 6, 4, and 2 were assigned to the worst
(13.1 feet) through the best (7.0 feet) ground resolutions
of the Scene-A and Scene-B GEMS. The ground-resolution
error score was the number of the Scene-C GEM judged equal
in ground resolution to the Scene-A (or B) GEM, minus the
number of -tile S. ene_,A~ .C B G N. ,, iThrreforep.ositi re, Y
error meaisb_,tn'at' tine S~cener Bj GtMan'd`'hega't?ive
error means it was judged to have worse ground resolution
than the Scene-C GEM.
Table 3 (Page 14) shows the results of the analysis
of variance of the errors of exposure-shift judgments. The
analysis showed there were statistically significant differ-
ences among the mean errors for the different exposure-shift
values (F = 8.63, p < .001), between the mean errors for
scenes (F = 68.13, p < .001), and a statistically signifi-
cant interaction between exposure shift and scenes (F = 3.11,
p < .05). In addition, the analysis of variance showed a
statistically significant difference between the mean errors,
on Trial 1 and 2 (F = 5.83, p < .05). The mean errors for
Trials 1 and 2 were .059 and .046 density units. Although
this decrease in error from Trial 1 to Trial 2 was statis-
tically significant, it was small and of no practical sig-
nificance.
Figure 5 (Page 13) shows the results associated with
all the statistically significant effects, except for that
of trials. The results showed that the errors in judging
exposure shifts were much larger for Scene A than for
Scene B. The mean error for Scene A was .108 density units,
and that for Scene B was .003 density units. On the
average then, the Scene-A GEMS were judged as denser than
the Scene-C GEMS, and the Scene-B GEMS were judged as the
same density as the Scene-C GEMS.
The change in mean error as a function of exposure
shift (the bowing of the two curves) is a commonly observed
Approved For Release 2002/06/17 : CIA-RDP78B04747A000700010023-9
 Approved For Release 2002/0 17=r:wc I DP78B04747A000700010023-9
TABLE 3
ANALYSIS OF VARIANCE OF ERRORS
OF EXPOSURE-SHIFT JUDGMENTS
SOURCE
df
MS
F
EXPOS
UR
E
SHIFT (A)
4
443.75
8.63***
GROUND
RE
SOLUTION (B)
3
33.00
1.23
SCENE
(
C)
I
27,417.50
68.13*
TRIALS
(D
)
1
361.50
5.83*
SUBJECT
S
(S)
TO
449.65
A X B
12
33.67
1.17
A X C
4
109.12
3.11*
A X D
4
29.00
1.22
B X C
3
6.00
0.30
B X D
3
6.00
0.28
C X D
1
106.50
2.05
A X B
X
C
12
26.29
1.05
A X B
X
D
12
30.79
1.36
A X C
X
D
4
42.50
2.54
B X C
X
D
3
10.83
0.44
A X B
X
C
X D
.12
33.21
1.33
A X S
40
51.39
B X S
30
26.75
C X S
10
402.40
D X S
10
62.00
A X B
X
S
120
28.78
A X C
X
S
40
35.09
A X D
X
S
40
.23.82
B X C
X
S
30
20.10
B X D
X
S
30
21.28
C X D
X
S
10
51.85
A X B
X
C
X S
120
25.01
A X B
X
D
X S
120
22.71
A X C
X
D
X S
40
16.72
B X C
X
D
X S
30
24.52
A X B
X
C
X D X S
120
24.92
***p