MULTIPLE IMAGE INTEGRATION VIEWER-PRINTER PHASE I FEASIBILITY STUDY

• Document Type: CREST
• Collection: NGA Records (Formerly NIMA)
• Document Number (FOIA) /ESDN (CREST): CIA-RDP78B04770A000200010048-8
• Release Decision: RIPPUB
• Original Classification: K
• Document Page Count: 63
• Document Creation Date: December 28, 2016
• Document Release Date: July 1, 2004
• Sequence Number: 48
• Publication Date: December 6, 1965
• Content Type: STUDY

STAT
I
Approved For Release 2004/07/29 : CIA-RDP78B04770
1
1
STAT
w
I
1
I
1
MULTIPLE IMAGE INTEGRATION
VIEWER -PRINTER
Phase 1-Feasibility Study
STAT
i
Declass Review by NGA. I
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
t AT
MULTIPLE IMAGE INTEGRATION
VIEWER-PRINTER
Phase 1-Feasibility Study
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
t
t
I
11 1
t
1
t
This report discusses the results of an introductory study of the problems of multiple image
integration and the details of the proposed design for an experimental three-input integration
viewer-printer. Automatic image correlation is used in the proposed equipment for rapid and
accurate registration of the three inputs. The outputs consist of a real-time electronic display
for monitoring purposes, and a high resolution, slow scan readout onto 4- by 5-inch film that will
reproduce image detail of up to 100 lines per millimeter on the original material.
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
e
r
t
t
2. Technical Considerations . . . . . . . . . . . . . . . . . . . . . 5
2.1 Image Registration . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Image Enhancement . . . . . . . . . . . . . . . . . . . . . . 19
2.3 Modulation Transfer Function . . . . . . . . . . . . . . . . . . 22
2.4 Comparison of Electronic and Optical Image Integration System . . . . . . . 29
3. General Description of Multiple Image Integration Viewer-Printer . . . . . . . 31
3.1 Control Panel . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Input Platen . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 Optical System . . . . . . . . . . . . . . . . . . . . . . . 34
3.4 Correlation System . . . . . . . . . . . . . . . . . . . . . . 38
3.5 Video Processing . . . . . . . . . . . . . . . . . . . . . . 41
3.6 System Performance . . . . . . . . . . . . . . . . . . . . . 52
4. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Appendix-Linear Superimposition of Multiple Images . . . . . . . . . . . . . 57
t
2-1 Maximum Tolerable Displacement Errors at 3 Percent Response . . . . . . . 7
2-2 Comparison of Electronic and Optical Image Integration Systems . . . . . . . 30
3-1 Input Magnification Ratios . . . . . . . . . . . . . . . . . . . . 52
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
1-1 Prime Transformations . . . . . . . . . . . . . . . . . . . . 2
1-2 Types of Terrain Relief Displacements . . . . . . . . . . . . . . . 3
2-1 Resultant Amplitude of Two Displaced Sine Waves . . . . . . . . . . . 6
2-2 Typical Modulation Transfer Functions for Aerial Photography . . . . . . . 7
2 -3 Effect of Image Displacement on Modulation Transfer Function . . . . . . . 8
2-4 Image Displacement Due to Tilt . . . . . . . . . . . . . . . . . . 9
2-5 Residual Second-Order Displacement Error Due to Tilt Angle . . . . . . . 10
2 -6 Residual Second-Order Distortion for 6-Inch Focal Length . . . . . . . . . 11
2-7 Residual Second-Order Distortion for 12-Inch Focal Length . . . . . . . . 12
2-8 Residual Second-Order Distortion for 24-Inch Focal Length . . . . . . . . 13
2-9 Residual Second-Order Distortion for 48-Inch Focal Length . . . . . . . . 14
2-10 Residual Second-Order Distortion for 100-Inch Focal Length . . . . . . . . 15
2-11 Residual Second-Order Distortion for 200-Inch Focal Length . . . . . . . . 16
2-12 Correlator Transfer Functions . . . . . . . . . . . . . . . . . . 18
2-13 Log Plot of Step Wedges in Sunlight and Shadow . . . . . . . . . . . . 23
2-14 Linear Plot of Step Wedges in Sunlight and Shadow . . . . . . . . . . . 23
2-15 Modulation Transfer Functions for Cathode-Ray Tube and Image Dissector,
Linear Plot . . . . . . . . . . . . . . . . 26
2-16 Modulation Transfer Functions for Cathode-Ray Tube and Image Dissector,
Logarithmic Plot . . . . . . . . . . . . . . . . . . . . . . 28
3-1 General Configuration of Multiple Image Integration Viewer-Printer . . . . . 32
3-2 Block Diagram of Multiple Image Integration Viewer-Printer . . . . . . . . 33
3-3 Control Panel . . . . . . . . . . . . . . . . . . . . . . . . 34
3 -4 Configuration of Input Platens . . . . . . . . . . . . . . . . . . 35
3-5 Diagram of Optical System . . . . . . . . . . . . . . . . . . . 37
3-6 Lissajous Scan Generator . . . . . . . . . . . . . . . . . . . . 39
3-7 One Channel of Video Correlator and Distortion Analyzer . . . . . . . . . 40
3-8 Block Diagram of Memory Unit . . . . . . . . . . . . . . . . . . 41
3-9 Block Diagram of Modulator Circuits . . . . . . . . . . . . . . . . 42
3-10 Block Diagram of a Single Image Dissector Channel . . . . . . . . . . . 43
3-11 Block Diagram of Variable Gamma Amplifier . . . . . . . . . . . . . 44
3-12 One Form of Nonlinear Amplifier . . . . . . . . . . . . . . . . . 44
3-13 Transfer Characteristic of Nonlinear Amplifier . . . . . . . . . . ... 45
3-14 Results of Step Voltage Signal Applied to Input of Gamma Amplifier . . . . . . 46
3-15 Video Signals . . . . . . . . . . . . . . . . . . . . . . . . 47
3-16 Schematic Diagram of Experimental Gamma Amplifier . . . . . . . . . . 49
3-17 Switching for Flicker Presentation . . . . . . . . . . . . . . . . . 51
3-18 Modulation Transfer Functions Referred to Image Dissector . . . . . . . . 53
3-19 Theoretical Modulation Transfer Function of Complete System . . . . . . . 54
e
1
r
t
VI I
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
1
t
Multiple image integration is the technique of utilizing several independent images of the
same object as an aid to photographic interpretation. The improvement obtained in reliability
and accuracy of interpretation may be due to any one or a combination of the following effects:
(1) improvement in the contrast of low contrast images, due to the coherent addition of image
detail and the noncoherent addition of grain or other random disturbances, (2) revelation of image
detail hidden in shadows or missing due to other causes, by the superimposition of two or more
inputs taken at different times of day or from different viewpoints, and (3) detection of changes
due to human activity or natural causes, e.g., seasonal variations.
One of the major problems in image integration is that of superimposing the multiple images
with sufficient accuracy to avoid loss of resolution, which would tend to offset any advantage gained
by the increase in contrast or visibility of detail. Methods of image integration using optical
superimposition have involved the use of multiple photographs taken at the same time from the
same viewpoint. The improvement obtained is due solely to the integration of photographic grain,
which theoretically improves image contrast by a factor of N112, where N is the number of photo-
graphs integrated. The problem of superimposition is reduced in this case simply to that of
physical alignment, since the inputs are all identical in scale and geometry.
To take advantage of all the benefits of image integration, however, photographs of different
scale taken from different viewpoints must be used. Superimposition then involves computation of
the geometrical transformations necessary to make the images congruent, followed by simultane-
ous implementation of these distortions on several images -a process similar to that required for
stereo viewing of convergent oblique or panoramic photography.
The required transformations can be computed either from a knowledge of the acquisition
parameters of the photography, or by simultaneous scanning of two or more inputs with subsequent
correlation of the imagery. The latter method is of far greater practical interest. At present,
electronic scanning is the only feasible method of point by point image correlation. The images to
be integrated are scanned by identical patterns, and the image density variations converted into
electrical video frequency waveforms. These waveforms are then compared in an electronic
correlator, and the image transformations necessary to render the images congruent are com-
puted. The image transformations can be implemented either optically or electronically by dis-.
torting the scanning patterns.
Image transformations can be analyzed into orders, depending on whether the image displace-
ment involved is independent of the coordinates x and y measured from the center of the viewed
area, or varies as x, x2, x3, etc.
Shifts in x and y which affect the entire image are zero-order transformations. There are
four first-order transformations in which image displacement is proportional to x or y, and six
second-order transformations in which image displacement is proportional to x2, y2, or xy.
These prime transformations are shown in Fig. 1-1.
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
Involving x
parallaxes
Involving y
parallaxes
y in x
(x skew)
xiny
(y skew)
Fig. 1-1 - Prime transformations
y2 in x
x2 in y
The transformations to be applied to two images to obtain congruence depend not only on the
viewpoint of the taking lens, but on the angular coverage of each image. Zero-order transforma-
tions (uniform displacements) are, of course, necessary whatever the image size. Similarly,
first-order transformations such as scale, rotation, and skew, in which image displacement is
proportional to the distance from the center of the viewed area, are necessary in all cases and
must be increasingly accurate as the number of resolution elements in the field of view is
increased.
Second-order distortions are present in every oblique photography. However, when only a
small area is viewed, as is generally the case with an image integration viewer, the second-order
distortion can be approximated with a straight line (first-order) correction. The superimposition
errors involved in this process have been analyzed and are reported in Section 2.1.2.
Third-order effects are present in panoramic photography in the form of S curve and cubic
scale distortion, but the magnitude of these effects has been found to be negligibly small when
measured over the area of photography normally viewed.
x in x
(x scale)
y in y
(y scale)
LLIIJ
x2 in x
y2 in y
1
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
t
1
Another form of image distortion must be considered in an image integration viewer, namely,
high order terrain relief displacements. First-order transformations already discussed will
correct for a constant terrain slope, as shown in Fig. 1-2a, while second-order transformations
will correct simple curvature, as shown in Fig. 1-2b. Terrain relief due to cultural objects shows
discontinuities in profile, as shown in Fig. 1-2c, corresponding to transformations which may reach
as high as 100th order. It is obviously impossible to correct these displacements by using a
systematic analysis, since the number of corrections to be applied would be astronomical. This
problem is discussed further in Section 2.1.3.
(a) First order
(b) Second order
Fig. 1-2 - Types of terrain relief displacements
(c) High order
The next question to be considered is how the required transformations can be applied to the
images. Of the ten possible first- and second-order transformations, only two (uniform scale
and rotation) can be removed by conventional optical means. Correction of x and y scale varia-
tions and skew distortion requires the use of anamorphic optics. Second-order distortion due to
obliquity can be corrected in an optical system by tilting the image and object planes with respect
to the lens axis. The use of these techniques results in a very elaborate and expensive optical
system which must be duplicated for each input channel of the image integration.
The problems associated with the use of optical systems for image superimposition lead to
the consideration of electronic imaging systems. Image transformations can be produced rela-
tively easily in an electronic system by changing the shape of the scanning pattern. Distortion of
any order, including high order relief displacement, is capable of being corrected in this way.
Because an electronic system must be used in any case to scan and correlate the images and
to compute the relative distortion, the use of electronic distortion removal techniques requires a
minimum of additional equipment. The main criticism leveled at electronic imaging systems is
their bandwidth restriction. There are, however, techniques by which this restriction can be
overcome to the point where the output imagery is suitable for photointerpretive use.
Electronic imaging systems using scanning reduce the two-dimensional input imagery to a
time varying video signal. Such a system is limited by three factors: (1) the size of the scanning
spot or aperture relative to the area scanned, (2) the bandwidth of the electronic video channel,
and (3) the size of the spot on the display tube relative to the area scanned. Factors 1 and 3
together determine the maximum number of resolution elements that can be reproduced, while
factor 2 determines the maximum rate at which they can be reproduced.
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
Image pickup and display tubes exist which can handle 9 x 106 image elements. This is equal
to the maximum number of image elements that can be seen by the eye (at 5 line pairs per milli-
meter in a 12- by 12-inch photograph at a distance of 10 inches. To present this information at
the normal rate of 30 frames per second would require a video bandwidth of 135 megacycles per
second, which is beyond the state of the art. However, the only reason for a frame rate of 30
frames per second is to produce the illusion of image motion without flicker. In an image integra-
tion viewer, we are dealing only with static imagery. If a readout system using image storage
is employed, the frame rate can be reduced to a few frames per second, and the full resolution
capability of the pickup and display devices can be utilized with a video bandwidth as low as 20
megacycles per second, which is readily achieved. It should be noted that the capability of handling
9 x 106 picture elements for electronic scanning devices approaches the performance of a lens.
While highly optimized lenses may exceed this performance, normally used lenses are capable of
reproducing about 107 image elements. Certain lenses in which compromises have to be made
may fall below this figure. So far as the pickup and display devices are concerned, therefore,
available electronic imaging systems come very close to matching optical systems. The overall
resolution of either type of system is of course reduced by virtue of the use of several compon-
ents in cascade.
Another aspect of image integration must be considered, namely, gamma control of the vari-
ous inputs in general, and shadow suppression in particular. The use of input material taken under
varying conditions of sunlight and in different seasons necessitates a correlation technique that
ignores illumination conditions and the presence of heavy shadows, and operates only on physical
objects. Since such objects are visible only by virtue of the illumination, this is not an easy task.
The main problem is the presence of shadows; the ARES correlation circuit locks onto shadows
very readily, which is quite acceptable when viewing stereo pairs obtained within a short time
interval. In the image integration viewer, it is necessary to reduce the contrast between shadows
and highlights for two reasons: (1) to enable images containing dissimilar shadow structure to be
correlated, and (2) to reveal shadow detail present in the images. The contrast can be reduced by
the use of nonlinear amplitude response (or gamma) in both the correlation and the viewing chan-
nels. A square root characteristic (gamma of 0.5) has been found effective in suppressing shad-
ows. It should be pointed out that the optimum video processing characteristics for correlation
and for viewing may not be the same. Therefore, we propose to use two independent gamma con-
trols in each channel, one to optimize correlation and the other to optimize viewing.
As a result of the study program, design parameters for a three-input image integration
viewer have been established. The instrument will accept three film chips of any size up to
9 i/2 by 9 i/2 inches. Correlation of the three inputs is achieved by electronic scanning using a
500-line, crossed diagonal scan. One input is designated as the master, and the geometry of the
other two inputs is transformed in sequence into congruence with the master input. The necessary
image transformations are stored in a memory. The scanned images are combined and displayed
on a viewing monitor, which enables the video processing in each channel to be optimized in terms
of tone reproduction and alignment.
When a satisfactory image has been built up, the system is switched to slow scan operation,
and the integrated image is recorded on film at high resolution. The scan time is about 1 second.
I
I
1
1
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
1
1
The first subject to be considered will be the required accuracy of registration. The distor-
tion due to camera tilt will be analyzed next, to determine whether second-order corrections will
be necessary in the image integration viewer. Finally, the effect of uncorrected relief parallax
will be considered.
2.1.1 Registration Accuracy
The effect of misregistration can be most easily evaluated by considering two sine-wave im-
ages, whose amplitudes, a, at any point, x, are
a1 = Ai sin (27rfix + (pi)
a2 = A2 sin (27rF2x + 'P2)
1
where Ai and A2 are the peak amplitudes, and fi and f2 are the spatial frequencies. If the two im-
ages contain the same spatial frequencies, and if they are added, the resultant, R = ai + a2, is
given by
R2 = (Ai cos q + A2 cos 02)2 + (Ai sin (p1 + A2 sin 02)2
R = [Ai + A2 + 2A1A2 cos ((P1 - 02)]112
The amplitude of the resulting image is therefore dependent on the phase angle, or displacement,
between the two images.
The phase difference can be expressed as
01 - 02 = 2vrfox
1
where Ox is the displacement between images. When Ox = 0, the two images are perfectly aligned
and the resultant amplitude is Ai + A2. When Ox = 1/4f, the sine waves are 90 degrees out of
phase, and the resultant amplitude is (Ai + A2)'2. When Ax = 1/2f, the sine waves are 180 degrees
out of phase, and the resultant amplitude is (MI - Ai)1i2. The variation in amplitude of the combined
sine wave is shown in Fig. 2-1 for Ai/A2 ratios of 1, 0.8, and 0.6. These curves can be used to de-
termine the loss in modulation transfer function (MTF) due to the misalignment of images.
Fig. 2-2 shows typical MTF's for an aerial camera lens (Itek 24-inch, f/3.5) and a high reso-
lution film (type 4404). The combined MTF shows that the response is down to 50 percent at
50 cycles per millimeter, and down to 3 percent at 300 cycles per millimeter. Fig. 2-3 shows the
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
MTF resulting from two inputs, each having the combined MTF of Fig. 2-2, when the displace-
ment, Ax, between them is equivalent to 1/4 wavelength and 1/2 wavelength, respectively, mea-
sured at a 300-cycle-per-millimeter spatial frequency. At less than a 1/4-wavelength image dis-
placement, the loss in MTF in most cases would be unimportant, whereas at image displacements
of 1/2 wavelength and greater, the loss in MTF is quite noticeable.
On this basis, we can establish the maximum tolerable displacement error at about 1/4f,
where f is the spatial frequency at which the MTF of the system has dropped to 3 percent. The
tolerable displacement error calculated on this basis is shown in Table 2-1.
1
1
t
Of /001"
?~
~_ Aj/A2 = 1
/
__Aj/A2 = 0.8
~._..A/A2 = 0.6
Displacement x Spatial Frequency, Ax x f
Fig. 2-1 - Resultant amplitude of two displaced sine waves
1
1
1
1
I
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
N
'~
,,-"
I
4404 film
=24-inch, f/3.5 lens
,
Combined response
15n 200 2
50 31
Spatial Frequency, cycles per millimeter
Fig. 2-2 - Typical modulation transfer functions for aerial photography
Table 2-1 - Maximum Tolerable Displacement
Errors at 3 Percent Response
System Resolution at
3 Percent Response,
lines per millimeter
Maximum Tolerable
Displacement Error,
microns
25
10.0
50
5.0
100
2.5
150
1.7
200
1.25
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
0
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
MTF of two inputs displaced 1/4
w
el
th
t 300
l
av
eng
a
cyc
es per
millimeter
MTF of two inputs displaced 1/2
wavelength at 300 cycles per
millimeter
MTF of inputs (from Fig. 2-2)
Spatial Frequency, cycles per millimeter
Fig. 2-3 - Effect of image displacement on modulation transfer function
2.1.2 Second-Order Distortions
Oblique and panoramic photography involve camera tilt angles that produce image distortion.
The basic types of image distortion are shown in Fig. 1-1. Zero- and first-order distortions, i.e.,
scale, skew, and rotation, must always be corrected in order to superimpose images of any size.
However, this is not the case with second-order distortions; the image displacements due to re-
sidual second-order distortion depend on the tilt angle and vary with the square of the y dimension
of the image area examined. It is important in the present case to determine whether or not sec-
ond-order corrections will be required in order to superimpose images to the required degree of
accuracy, since this will be a major factor in deciding the form of the final equipment.
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
t
i
I
The required superimposition accuracies for materials of various resolutions were deter-
mined in Section 2.1.1. If second-order corrections are necessary, there is a strong argument for
using an electronic distortion removing system, where second-order corrections can be made by
changing the shape of the scanning pattern. This is achieved far more easily than the correspond-
ing correction in an optical system, where the use of tilted planes is required.
The scale of any image on a tilted photograph can be expressed as
S_F - ysint
H -h
where t = tilt angle
y = distance of image from the isocenter measured in the direction of the tilt
H = flying height of the camera
h = ground elevation of the object
F = focal length of the system used
The rate of change of scale is therefore
dS _ sin t
dy H - h
Dt (F/sin t) - y
Tilt can also be expressed in terms of the displacement of the image (Fig. 2-4) by
Since -y is normally small compared with F/sin t,
Dt=
1
t
t
sinty2
F
Fig. 2-4 - Image displacement due to tilt
I Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
The image displacement is therefore of a parabolic form as shown in Fig. 2-5. An automatic
registration system with only first-order capability will provide a linear approximation to this
curve, as shown by the line AA. The residual peak to peak error is 2e. It can easily be shown that
for the curve D = (sin t/F) y2, the residual second-order error is
sin t
e 8F (y2 - Yi)2
where yi and y2 define the edges of the viewed area in the y direction.
d_sin t 2
f y
sin t
e 8f (Y2 - Yi)2
D2
t
r
t
I
Fig. 2-5 - Residual second-order displacement error due to tilt angle
I
Residual second-order errors have been computed for tilt angles of 7.5, 15, 30, and 45 degrees,
with camera focal lengths of 6 inches (150 millimeters), 12 inches (300 millimeters), 24 inches
(600 millimeters), 48 inches (1,200 millimiters), 100 inches (2,500 millimeters), and 200 inches
(5,000 millimeters). The results are shown in Figs. 2-6 through 6-11. In each of these figures,
the maximum permissible superimposition errors for images of resolutions of from 25 to 200 lines
per millimeter have been indicated, enabling the effect of second-order distortion on the final im-
age to be estimated. Figs. 2-6 through 2-11 show that, in general, second-order corrections are
essential, especially at large tilt angles and with high resolution imagery.
t
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 t
1
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
Tilt 4
5?
30?
15?
7.5?
25 L/
mm
50 L/
mm
Maximum
displacement
error for
stated
f 100 L/
mm
resolution
-a-200 L/mm
y Dimension of Image, millimeters
Fig. 2-6 - Residual second-order distortion for 6-inch focal length
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
Tilt 45?
30?
15?
?
7.5
5 L/mm%
50 L/mm
Maximum
displacement
error for
stated
.100 L/mm
resolution
-a-200 L/mm
y Dimension of Image, millimeters
Fig. 2-7 - Residual second order distortion for 12-inch focal length
t
t
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 t
t
t
t
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
Tilt 45?
30?
150
7.5?
25 L/m
m
50 L/mm
Maximu
displace
error f
o
stated
f 100 L/mm
resoluti
200 L/mm
y Dimension of Image, millimeters
Fig. 2-8 - Residual second-order distortion for 24-inch focal length
m
ment
r
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
Tilt 4 50
?
30
15?
7
.5?
X25 L/
mm
--.0.50L/
Maxi
mm
displa
error
stated
resol
.,.100 L/
mm
X200 L/
mm
y Dimension of Image, millimeters
Fig. 2-9 - Residual second order distortion for 48-inch focal length
mum
cement
for
ution
1
t
I
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
t
t
t
t
Tilt 45?
300
150
7.5?
25 L/mm
mw.
Maximum
50 L/mm
displace-
ment
error for
stated
resolution
100 L/mm
200 L/ mm
1
t
y Dimension of Image, millimeters
Fig. 2-10 - Residual second-order distortion for 100-inch focal length
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
Tilt 45'-
300
15?
7.5?
25 L/mm -~
Maximu
m
50 L/
m
m-..P-
displace-
ment
error for
stated
resolution
100 L/m
m-.w.-
200 L/ mm ~-
I I I I /~
/
I I
I I I I -Lj
y Dimension of Image, millimeters
Fig. 2-11 - Residual second-order distortion for 200-inch focal length
t
1
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
IAT
t
t
1
2.1.3 High Order Corrections
The high order image distortion produced by relief displacement presents a problem in image
integration. Obviously, the differences in imagery due to viewpoint must be eliminated or com-
pensated for, in order to obtain the full benefits of image integration. There are two approaches
to the problem (1) eliminating relief displacements by making each element of imagery congruent
with the corresponding element in other inputs, and (2) using stereo viewing to take advantage of
the information present in relief displacements.
If the images to be integrated have been taken from similar viewpoints, or if the vertical relief
in the viewed object is small, the problem is considerably simplified; in these cases it will be pos-
sible to directly integrate several photographs without being concerned with high order corrections.
High order displacements can be eliminated on a point by point basis in an optical system, the
process being similar to the production of an orthophotograph. This is, however, a slow process
and not suitable for photointerpretive use. An electronic scanning system is potentially capable of
removing high order image displacements almost instantaneously for direct viewing. Although this
is not yet a practical reality, development of the ARES viewer to achieve this capability is being
vigorously pursued atE= The best immediate solution for the integration of multiple images con-
taining considerable vertical relief obtained from several different viewpoints appears to be to sep-
arate the input material into groups, such that the variation in viewpoint in a single group is mini-
mized. Each group can then be integrated directly. The integrated groups can then be taken in
pairs and viewed stereoscopically; each pair may involve a different stereo base line and thus give
a different view of the target. This approach seems to make the best possible use of the different
capabilities of a multiple image integration device and a stereo viewer. Ultimately, the develop-
ment of an image integrator with high order transformation capability will enable the viewpoints to
be exactly compensated, so that a single integrated photograph, or a single stereo pair, will con-
tain all the information present in the various inputs. The proposed electronic viewer will have
this add-on capability for high order image transformation, and will not become obsolescent.
2.1.4 ARES Correlator Performance
Having determined the registration accuracy requirements for image integration, we will now
consider the performance of the current ARES correlator.
The ARES correlator and distortion analyzer produce their outputs by averaging the errors
over the whole viewed area. The accuracy with which distortion errors are computed is not there-
fore limited by the number of lines in the scan, since there is a considerable integration effect
present. Typical correlator transfer functions are shown in Fig. 2-12. In the process of image
registration, the low frequency correlation channel comes into operation first, since this has the
widest range of operation.
As the image parallax is reduced, the displacement comes within the range of the higher fre-
quency channels. The lower frequency channels are successively eliminated in order to reduce
their noise contribution. The final accuracy of the system therefore depends on the transfer char-
acteristic of the highest frequency correlator channel, the loop gain of the system, and the cor-
relator noise. The correlator transfer curve has a slope of 2 volts per 1 percent of raster width.
For an image 0.25 inch square, which is the smallest viewed area in the proposed design, the cor-
relator gain is 800 volts per inch, or 0.03 volt per micron. If sufficient loop gain is provided, the
limit on registration accuracy is set by the noise level at the correlator input. The noise level
varies with the type of imagery being scanned, but is of the order of 0.1 volt. The minimum dis -
placement error that can be sensed is therefore approximately 0.1/0.03 or about 3 microns. This
error is consistent with the displacement error that will allow a resolution of about 100 lines per
millimeter in the integrated image, referred to the input material, which was determined to be
2.5 microns in Section 2.1.1 (see Table 2-1). This 3-micron displacement error corresponds to a
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Parallax, percent of raster width
(a) Channel A - 100 to 215 kc/sec, fc = 147 kc/sec
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
1.6
0.8
D 3 6
9
12 V
(b) Channel B - 215 to 460 kc/sec, fc = 316 kc/sec
Parallax, percent of raster width
(c) Channel C - 460 to 1,000 kc/sec, fc = 680 kc/sec
Fig. 2-12 - Correlator transfer functions (raster width is 18 millimeters)
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
t
t
 t
t
1
t
1
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
spatial frequency phase shift of 90 degrees at about 83 lines per millimeter. The amplitude of the
resulting MTF of three combined images will be lower than this by a factor of approximately
1/31/'2 or 0.58. In computing the effect of image misalignment on the final image, it must be kept
in mind that an image integration viewer cannot increase the detail seen in high contrast images.
The purpose of integration is to improve the signal to noise ratio, and therefore the visibility, of
low contrast detail. This increase in visibility is necessarily accompanied by a small decrease in
the limiting resolution of the system due to small but finite image misregistration.
Subjectively, the quality of reproduction produced by a television system is described by the
term "photographic quality." The comparison between electronic display quality and photographic
quality is rated in terms of the photographic system because of its higher performance. Those
factors that are judged in the comparison are tonal quality, signal to noise ratio, and picture
sharpness. These factors are interrelated in electronic systems, and tradeoffs are required to
produce an image with satisfactory characteristics on the display system.
The control of the relationship of the reproduced gray scale to the luminance of the original
subject is of importance in achieving excellent tonal quality. According to the Weber-Fechner law,
the eye perceives changes in luminance in a logarithmic relationship (i.e., an increase in stimulus,
dB, is proportional to the existing stimulus, B, or, for equal changes, AB, SB/B is constant) there-
fore, the transfer function of the system is generally plotted in terms of the logarithm of the input
luminance to the logarithm of the output luminance of the system.* The gradient of the system
describes the contrast of the reproduction compared to that of the original subject, and is the slope
at any point of the gray scale dynamic transfer function (GTF) when plotted on log-log paper.
The maximum slope, or gradient, is termed the "gamma." A maximum slope of 45 degrees
is equivalent to a gamma of 1, and indicates a 1:1 correspondence between input and output at that
point. The ratio of log input to log output is thus proportional to the value of gamma.
The eye sees a uniform decrease in contrast between steps in a gray scale as the gamma is
decreased below 1. However, in linear coordinates, the black end of the GTF is expanded while
the highlight, or white end, is compressed. For gammas greater than 1, the opposite effect takes
place, i.e., the white end of the gray scale is expanded and the black end compressed. In the latter
case, the eye sees the contrast between steps increase uniformly.
Gamma is measured at the point of maximum slope and reflects the gradient maximum. Be-
cause of the nature of almost all gray scale reproducing systems, there is compression at both
ends of the scale. The effect of this S distortion on reproducing systems is that for a constant
GTF ratio, the gradient is increased for the central or middle grays and compressed for the high-
light and shadow areas. The average gradient is therefore always less than the gamma of the sys-
tem. It is the increase in overall average gradient that marks the reproduction as being inferior
in quality to its original.
In photography, and particularly in the field of color reproduction, this factor is well known.
Techniques such as highlight masking are used to help overcome the degradation of successive
generations of reproduction.
Electronic systems are very flexible in handling this problem. Amplification of the video sig-
nal with nonlinearities inverse to those nonlinearities present in the original material and in the
* Since d[ln (B/Bmax)] = dB/B.
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
video channel can correct for gray scale distortions. In fact, it is possible in some cases to im-
prove on the original by removing tonal distortions caused by improper processing or exposure.
The basic limitation to the use of the above techniques is the noise component of the signal,
whether it be derived from the original photography or generated in the video circuits used in re-
producing the imagery.
The overall dynamic transfer functions of both optical and electronic systems are degraded by
noise. The signal to noise ratio for an image reproducing system is a measure of the maximum to
minimum contrast, or contrast ratio, that can be handled. In optical systems, the noise is caused
essentially by stray or flare light which reduces the dynamic range of the system. Electronic sys-
tems suffer from this same problem of unwanted light, as well as from noise generated by the
electronic components of the system.
These factors limit the amount of correction that can be supplied by electronic gamma cor-
rection. For rooting functions, as larger contrast ratios are handled, the gain ratio increases for
the blacker parts of the video signal. This may involve high gains in the dark gray region, and
may thereby enhance defects resulting from (1) noise generated in the transparency scanning sys-
tem, (2) stray light (or flare) in lens systems, and (3) inaccurate black level adjustment.
Therefore, in correcting for the nonlinearities, as greater contrast ratios are accommodated
a point is reached where the disadvantages begin to equal the advantages. This point must be de-
termined empirically, and will vary with different inputs. Therefore, the amount of gamma cor-
rection must be controlled by the operator.
Gamma Correction Circuitry
In a video system, part of the characteristic is electrical, and this part of the overall GTF is
expressed in terms of the ratio of the electrical input and output signals (voltage or current). A
video signal which has been transferred through an ac coupling will be zero for some intermediate
luminance value. When the peak to peak video signal is reduced to zero, there remains a dc com-
ponent, either zero or some other value, which represents the average value of the signal. If the
scene being transmitted is almost all white, the average is near the white peaks; if the scene is
almost all black, the average is in the dark gray portion of the video signal.
Before the signal can be applied to a nonlinear amplifier, it is necessary to set a certain value
of the signal at a fixed dc point, so that the corresponding luminance value will always appear on
the nonlinear characteristic at the correct point.
Experimental work was conducted on the design of an amplifier for insertion in the ARES
video channel for the correction of nonlinearities. This work was divided into three phases:
1. The establishment of a black level in the video signal by the addition of a blanking pulse
to the flying spot scanner tubes
2. The development of an amplifier to drive the nonlinear circuitry and clamping system
(keyed clamp), to establish an operating point at the black level during the clamping period
3. The development of a nonlinear amplifier with single knob control of the nonlinearity.
Two breadboard gamma amplifiers were constructed and installed in the ARES for evaluation.
In addition, a circuit was breadboarded for electronic switching between two video inputs at vari-
ous rates without flicker. The arrangement was such that it was possible to switch each of the two
ARES channels to a straightthrough condition, or to insert the gamma amplifier. When inserted,
the gamma amplifier output fed both the correlation channel and the video driver for the display
tube.
20
1
r
r
I
t
I
I
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
The purpose of our tests was to demonstrate the application of manual control of gamma to
enhancement of the visual display, and also to determine what effect a change in gamma would have
on correlation.
Blanking signals were fed to the cathodes of the two scanning cathode-ray tubes by means of a
blanking generator. The blanking signal was "folded" around the edge of the raster and its ampli-
tube adjusted to cut off the cathode-ray tube light for 2 microseconds. The blanking pulse then ap-
peared on the video output of the photomultiplier as a pulse representing black (absence of light),
which could be used as a reference in the do restoration circuits at the gamma amplifier and also
at the grid of the viewing cathode-ray tube.
The gamma amplifier used a keyed clamp circuit to establish a constant operating point for
the nonlinear amplification of the video signal. At the cathode-ray tube, it was found that the black
setting circuitry of the ARES clamped this point, since it was the blackest point in the picture. It
was noted that the introduction of the blanking pulse to the ARES circuitry improved the appear-
ance of the black portions of the viewed picture, permitting observation of detail in the blacker
portions that otherwise was lost due to black compression. It was also noted that the time con-
stants of the ARES circuitry, when photographs containing large, massive areas of black were
used, caused the black setting operation to take place on these areas rather than on the blanking
pulse. As a result of these observations, it is recommended that keyed clamps be used at all
points requiring accurate black level setting.
The circuit described above operates by detecting the blackest part of the picture and setting
a constant cathode-ray beam current at this point representing black. The circuit works very well,
but the results can be confusing. For example, when no photograph is in the viewing part of the
ARES, the darkest part of the video signal is created by negative peaks of the phosphor grain pat-
tern and the pattern caused by burning of the phosphor, which is a normal effect as the tube ages.
The black level setter adjusted these negative peaks to represent subjective black as referred to
the peak light output. The white excursions of the noise (grain and burn) signal were expanded by
the automatic video gain to full peak output. Normally the grain pattern of a scanning tube phos-
phor has an rms value of less than 10 percent near the average signal level. However, this circuit
expanded the noise signal to approximately 95 percent.
This form of image control presents the possibility of severe tonal distortion for certain types
of photography, in particular, that which has no true subjective blacks. By using an established
black reference point throughout the image integration viewer system, this form of tonal distortion
will be avoided.
2.2.2 Shadow Suppression
The automatic registration of several aerial photographs of the same area taken under dif-
ferent conditions of illumination is sometimes difficult because of the presence of shadows or
specular reflections. Shadows are a dominant feature of photographs taken with sunlight falling
on the area from a low angle. Specular reflections, while less common, cause obliteration of
significant detail due to saturation of the photographic material. If the shadows in the several
photographs are in different positions due to a change in the position of the illumination source,
or are present in only a few of the inputs, correlation is difficult to achieve. A great improve-
ment in the operation of automatic registration circuitry can be realized if the effects of shadows
can be eliminated.
A better understanding of the problem may be had from the following considerations. Assume
that we have two reflection density step wedges, each of the ten steps having a density greater by
0.15 than the preceding step, with step 1 being of zero density and assumed to reflect 100 percent
of the light. We place the two step wedges in the field to be photographed. At the instant we take
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
the photograph, a cloud obscures the illumination on one of the two tablets, reducing the illumi-
nation on it to be 8 percent of the illumination falling on the other. A second photograph is made
at a later time, when the cloud has passed. Assume also that the photographic material used in
making the pictures has no compression on the toe or shoulder of the density-log exposure char-
acteristic curve.
If we then plot the density of each step on the developed photograph, our plot shows that each
step when plotted against its density (Fig. 2-13) has a constant height relative to the other steps.
This is how the eye sees the photograph. The step value, in terms of density, is the same whether
the step is in full sunlight or in shadow.
However, when the photograph is placed in the flying spot system, the video signal will be as
shown in Fig. 2-14. Since the photomultiplier of the flying spot system responds linearly, each
step will be seen to decrease proportionally by a constant percentage of the previous step.
The shadowed step wedge is seen to occupy only 8 percent of the available video channel
amplitude, and the brightest step is step 1. We can, by increasing the gain of the amplifier, bring
shadowed step 1 up to 100 percent, or sunlight, illumination, thereby eliminating the effect of the
shadow. However, if we must see some of the area outside the shadow, we cannot do this.
The other alternative is to take the logarithm of the signal shown in Fig. 2-14. On the output
of the logarithmic amplifier, the steps will have equal increments and will appear as in Fig. 2-13.
If, in the correlator, we differentiate the signal or pass it through abandpass filter, thereby re-
moving the dc component and the lower frequencies, we should be able to improve the possibilities
for automatic registration.
The methods outlined above, namely (1) to look entirely within the shadow and raise the video
gain, and (2) to take the logarithm of the signal and differentiate, are subject, of course, to the
realities of the situation. We do not have photographic materials without compression in the toe
and shoulder regions of the D-log E curve, so that even the logarithmic approach will not produce
steps of equal amplitude for the shadow and highlight detail. Nevertheless, this will improve the
situation and the number of photographs that can be automatically registered should be increased.
It has been our experience in the experimental application of nonlinear amplification in the
correlation channel of the ARES that there is a variable in the nature of the nonlinearity that is
necessary for achieving optimum conditions for automatic correlation. On certain subjects,
stretching of the blacker portions of the video signal seemed to improve correlation, while for
other photographs, stretching of the white was found to be more beneficial. For this reason, it
is our recommendation that the operator be provided with a means of selecting the characteristic
to be inserted in the correlation channel. This will be accomplished by a switch which will select
one of the following functions of the video signal to be inserted in the correlation channel:
1. Square law-to emphasize the whiter portions of the signal
2. Linear-the normal mode as now used
3. Logarithmic-to emphasize the blacker portion of the signal.
The prediction of performance characteristics of both the optical and electronic systems is
computed in the same manner. The response of optical and electronic systems combined can be
calculated to arrive at a measure of total performance in resolution capability.
In calculating the response of the system, MTF data for each component of the system are
required. Modulation transfer, sometimes called sine-wave response or contrast transmission
function, relates the output amplitude of a sine-wave signal to the input amplitude as a function of
22
t
t
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved-For Release 2004107/29: ;TA-
ERRATA
On p. 22 and in Fig. 2-13 and 2-14 For:8 percent;Read:12.5 percent
(Sunlight, 100 percent illumination
2.4 L
0
Shadow, 8 percent illumination
Step Number
Fig. 2-13 - Log plot of step wedges in sunlight and shadow
Sunlight, 100 percent illumination
Shadow, 8 percent illumination
0 2 4 6 8 10
Step Number
Fig. 2-14 - Linear plot of step wedges in sunlight and shadow
23
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 frequency. Simply expressed, it is the frequency response of the component or system. When the
MTF of each component of a system is known, the system MTF can be obtained by finding the
product of all responses for each frequency.
The term "frequency" as used in this context relates to the physical size of the picture ele-
ment. The higher the frequency, the finer the size of the element. This is a crude means of ex-
pressing the differences between the concept of "spatial" or "time" domain (the former applying
to optical systems, the latter to electronic or scanning systems), and "frequency" domain.
The resolution of optical systems is measured in line pairs or cycles per unit length. In
electronic video systems, the number of scanning lines is fixed, and therefore, vertical resolution
is also fixed. The size of the electronic display system may vary from one installation to the next.
Therefore, the concept of frequency response of the video system must be related to a common
factor, i.e., the height of the picture. Thus, the expression, "a 500-line picture," means that each
picture element is 1/500 of the picture height, regardless of the actual dimensions of the display.
Another important difference between optical and video resolution measurements is that the optical
line pair is equal to two television (TV) lines. Therefore, in the display of a 3,000-TV-line picture
in a raster 3 inches high, the ultimate resolution is 1,000 TV lines per inch, or 500 line pairs per
inch.
It should be noted that the resolution referred to here is along the television line. The reso-
lution in the vertical direction, i.e., at right angles to the direction of the scan line, is modified by
the Kell factor. Determination of vertical resolution is a sampling process. The Kell factor
states the probability that given a sine-wave test target whose frequency in TV lines is equal to
the number of TV lines in the scanning pattern, and which is oriented such that each half-cycle is
scanned by one TV line, (1) the output will not be zero, as would be the case if the scanning line
were to traverse the test target along the zero crossing, and (2) the output will not be a maximum,
as would be the case if one line traversed the peak of the positive cycle and the next the peak of the
negative cycle. The actual probability is the root mean square of the two extremes, and is there-
fore generally considered to be equal to 0.7; hence, the average vertical resolution is 0.7 times the
number of scanning lines. This means that to obtain vertical resolution equal to 3,000 horizontal
TV lines, 4,200 active TV scanning lines are required. The empirical value of the Kell factor, as
measured by a number of experimenters, has varied from 0.5 to 0.89, and depends on the type of
display, the actual size of the video camera scanning spot, and other factors.
2.3.1 Flying Spot Scanning Systems
. It has become the accepted procedure when generating a video signal from photographic trans-
parencies to use a flying spot system. The system consists of a cathode-ray tube upon whose
phosphor screen is generated a rapidly moving light spot, tracing out a raster by the action of an
electronic beam. The light from the raster is imaged by a lens onto a photographic transparency.
The light modulated by the transparency is collected by a field lens or condenser system and
transmitted to a photomultiplier whose output represents the instantaneous value of transmittance
of the photographic material as a function of the position of the spot in the cathode-ray tube raster.
It is generally accepted that resolution and tonal qualities of the video signal generated by flying
spot systems are superior to those produced by other forms of television systems, such as those
which use the vidicon and the image orthicon.
The limitations of the flying spot system are imposed by the spot size of the scanning beam
and the persistence of the phosphor, the latter limitation being the greater. The fastest phosphors
decay to the 10 percent point in about 0.1 microsecond. This limits the practical bandwidth obtain-
able from flying spot systems to the order of 15 megacycles.
The requirement for a flicker-free image requires a frame rate of 30 frames per second. If,
in meeting the requirements for high resolution printout, this requirement of flickerless display
24
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
can be ignored, it is possible by reducing the frame rate to transmit a larger number of picture
elements within the 15-megacycle limitation imposed by the decay time of the phosphor. With a
slow scan system, the persistence factor is eliminated and the limitation on resolution and con-
trast is imposed by the spot size and the noise generated in the photomultiplier tube.
The best combination of phosphor efficiency and phototube sensitivity is found in the P-16
phosphor, which peaks in the ultraviolet region at 370 millimicrons, in conjunction with an ultra-
violet sensitive photomultiplier. This allows the best possible beam current reduction in obtain-
ing the best signal to noise ratio. The limiting resolution of such a system for commercial tele-
vision application is 800 lines per frame, with signal to noise ratios close to 100:1 (40 db).
Image Dissector
The image dissector surpasses all known television pickup devices in resolution. All other
camera tubes and flying spot scanner systems use a cathode-ray beam to perform the scanning
operation or to illuminate the flying spot. The definition of a cathode-ray beam is never as sharp
and exact as the direct imaging of an electronic image, as in the dissector or in an image con-
verter tube. Measurements on an image converter with a transparent phosphor showed a resolu-
tion of 140 line pairs per millimeter (equivalent to 280 TV lines per millimeter). Comparable
resolution could be achieved with a image dissector if the mechanical aperture were fine enough.*
Fig. 2-15 shows the MTF for an image dissector and a 1-mil cathode-ray tube.
The limitation on the use of the image dissector is not on obtaining sufficiently fine mechani-
cal apertures. An aperture of 0.0003 inch in a tube will give sufficient resolution to resolve
3,000 TV lines in a 1-inch-high raster. With conventional dissector techniques at high scanning
rates, this is not achievable, due to signal to noise limitations connected with the amount of photo-
cathode illumination permitted before deleterious effects set in.
The signal to noise ratio of the dissector for a constant fixed scanning rate is proportional to
the square of the linear dimension of the aperture. As a result, a dissector designed for com-
mercial television requires an aperture of 1/500 of the picture height, or for a 1-inch raster, an
aperture of 0.002 inch. Compared to a similar tube having a 1/3-mil aperture, the signal to noise
ratio will be (2 x 10-3/3 x 10-4)2 = 4 x 10-s/9 x 10-$ = 44.
The signal to noise ratio also varies as the square root of the sweep time. For a fixed photo- STAT
cathode loading, doubling the sweep time will increase the signal to noise ratio by 40 percent.
An image dissector tube, WL-23111, manufactured by the
has a limiting resolution of 3,200 TV lines. Its photocathode can be loaded
with illumination equal to a current density of 4 microamperes. For commercial TV line scans
of 63 microseconds, the tube's peak to peak signal to rms noise ratio is 0.4. This, of course, is
impractical and is the reason that the image dissector has not been universally adopted for stan-
dard television systems.
With conventional illumination, the WL-23111 image dissector must be used with slow rates
in order to produce signals sufficiently greater than the noise component of the photocathode cur-
rent. Since the signal to noise ratio varies as the square root of the scan time, a scan time of
0.6 second per horizontal line is necessary to obtain a ratio of 40:1.
For a printout system utilizing the maximum resolution of the dissector, assuming a 0.7 Kell
factor, 4,200 TV lines will be required to give equal horizontal and vertical resolution and to
mask line structure in the output print. The product of 4,200 lines and 0.6 second is 2,520 seconds
for one frame time, or approximately 43 minutes.
* G. Papp, On a Novel Application of the Image Dissector, J. SMPTE, 74:782 (Sept 1965).
25
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
TV Lines Per Frame Height
4,000
5,000
6,000
23 11 image dissector,
0.3-mil aperture, 1-inch raster
1-mil spot, 3-inch raster
on crt
1,000
3,000
2,000
1,000
3,000
1,500
2,000
Fig. 2-15 - Modulation transfer functions for cathode-ray tube and image
dissector, linear plot
"Follow Spot" Technique
Our discussion of the image dissector has led to the conclusion that the tube is impractical
for conventional scan techniques because of the signal to noise ratio imposed by the illumination
limitation. However, by using a flying spot cathode-ray tube to illuminate only the small area
being directly scanned on the dissector photocathode tube, in much the manner that a "follow spot"
is used in the theater to follow an actor as he moves across the stage, the signal to noise ratio of
the dissector can be improved enormously. We still meet the limitation of a 4-microampere total
cathode current, but because of the short duty cycle of illumination per picture element, the major
portion of this current now reaches the scanning aperture.
The follow spot from the flying spot tube can be considerably larger than the picture element
of the dissector tube, thereby minimizing tracking problems. The persistence of phosphor and the
spot size of the cathode-ray tube do not affect the resolution of the image dissector.
This technique is currently used for high resolution displays. Its use is recommended for the
image integration viewer and printout system, although the complexity of the combined image dis-
sector and flying spot system is much greater because of the requirements for geometric scan
correction and electronic zoom. However, we believe that these complexities are a small price to
pay for the increased sharpness and resolution that can be obtained from the image dissector.
26
I
1
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
1
2.3.2 High Resolution Printout System
Photographic records of the superimposed imagery will be made in a module component of the
image integration viewer-the hard copy printout system. The basic components of the printout
station are a cathode-ray tube with high resolution display capabilities, and a P-11 phosphor im-
aged through a lens onto the image plane. A Graflok back will be provided at the image plane to
permit the use of Polaroid materials, as well as standard 4- by 5-inch cut film and roll film.
The operation of the system will be initiated by a control on the viewer control panel, after
the operator has made the necessary adjustments to the imagery as viewed on the system monitor.
As discussed elsewhere, the monitor operates to provide a flicker-free display for direct viewing.
When the printout operate button is depressed, the system shifts from a 30-frame operation to a
slow scan mode. The linearities and geometrical transformations needed to achieve correct reg-
istration of the several inputs are stored in the system memory, which controls the scanning at
each of the input stations.
Readout Cathode-Ray Tube
It is our objective in the design of the printout system to maintain the inherent resolution of
the video signal derived from the image dissectors. To this end, the selection of the cathode-ray
tube for recording the video image photographically plays an important part.
The cathode-ray tube parameters that determine its suitability for use in a photographic re-
cording system are (1) the photoactinic light output, (2) the spot size, or ability to reproduce fine
detail with good contrast, and (3) the noise or granularity of the phosphor.
The actinic light output is the result of the spectral characteristics of the phosphor, and the
photographic efficiency is rated on the basis of the density obtained for a given exposure. Nor-
mally, the spectral sensitivity of the film must match the spectral sensitivity of the phosphor light
output, to ensure maximum efficiency. Persistence of the phosphor is important in systems where
change of information takes place in successive scans, in which case a carryover of the previous
image would be detrimental. In the proposed system with a slow scan output, the persistence will im-
prove the photoactinic efficiency. The phosphor that most closely meets the requirements for the
printout system is P-11, a blue zinc sulfide. Its peak spectral output is at 460 millimicrons. Its
average persistence is 60 microseconds at the 10 percent point; the actual decay time is a function
of the beam current. A number of manufacturers are able to supply high resolution tubes with a
fine grain version of this phosphor.
The resolution on the P-11 cathode-ray tube is dependent on the spot size formed by the elec-
tron gun on the phosphor screen. The spot dimensions are determined by the design of the elec-
tron optics of the gun and also on the amount of spreading of light that occurs in the phosphor
screen material. Tubes with a nominal 1-mil spot width, measured at 60 percent of the luminance
peak, are available in tubes with a 41/ -inch optical quality screen. A raster 3 inches square will
fit within this circle. If the spot shape istobe Gaussian, its transform to frequency response is also
Gaussian, as expressed by
F = exp (-27r2f2o2)
where F is the percent modulation for a spatial frequency, f, and the spot width measured at
60 percent of peak output is 2Q.
The results of a computation using this relationship are plotted in Fig. 2-16 for a raster of
3 -inch frame height, and a Gaussian spot width of 1 mil. The 3 -inch raster and 1-mil spot size
are compatible with the 4,200 active line scan proposed for printout. The frequency response of
the WL-23111 image dissector is plotted in terms of its frame height, and the combined result of
27
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
the dissector and the output cathode-ray tube is computed and plotted. It is seen that the 50 per-
cent response point for the two is approximately 850 line pairs per frame height. The limiting
response (3 percent) is approximately 1,400 line pairs.
2,000
4,000
101000
20,000
WL-23111 image dissector
1-mil Gaussian spot, 3-inch-high
raster on cathode-ray tube
Combined response
1 mil at 60 percent
amplitude
100 percent
amplitude
1,000
5,000
10,000
1,000
2,000
Fig. 2-16 - Modulation transfer functions for cathode-ray tube and image
dissector, logarithmic plot
The magnification required for printout is dictated by the film size, usually 4 by 5 inches, or
if roll film is used, 21/2 by 3/ inches on 70-millimeter film. It is recommended that the output
device operate at a 1:1 magnification for all film sizes.
For 1:1 magnification, 850 line pairs in 3 inches corresponds to 12 lines per millimeter on
the film plane. The MTF for the film and imaging lens show almost full response at this point;
thus, the film image should have essentially the same response as appears on the face of the
cathode-ray tube.
1
1
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
Twelve lines per millimeter is somewhat finer detail than the eye can resolve at a 10-inch
viewing distance. A magnification of 2x will bring the finest detail within the resolving capability
of most observers.
It should be recalled that we have not been discussing the limiting resolution of the system as
is normally done in optical and photographic systems, but rather the 3 db down or 50 percent re-
sponse point. This is compatible with measurements made on electronic equipment.
2.4 COMPARISON OF ELECTRONIC AND OPTICAL IMAGE
INTEGRATION SYSTEMS
A comparison of the features of electronic and optical image integration systems is given in
Table 2-2. The features examined are those essential to an integration viewer, i.e., sensing of
parallax, transformation capability for handling inputs of varying geometry, gray scale process-
ing, resolution, and type of output.
This is not intended as a general comparison between electronic and optical viewers. In a
scanning viewer, for example, the resolution of an electronic system would be restricted by the
necessity of presenting the information at a high frame rate. A high frame rate is not required in
an integration viewer, since the imagery is essentially static.
In summary, it can be stated that for the present purpose an electronic system has the advan-
tage of a far greater flexibility in image transformation and gray scale processing. The limiting
resolution of the hard copy image on film based on a crt spot size of 1.5 mils would correspond
to about 100 lines per millimeter over an area of 6 by 6 millimeters at the input film. The total
number of image elements is about 1.6 x 106. Maximum overall magnification will be 12x, giving
an output image area of 3 by 3 inches, with a limiting resolution of about 8 lines per millimeter.
An optical image integration system would be limited as to the types of materials that could be
handled, as regards image geometry and film processing (e.g., negative material could not be
mixed with positive). Our experience with the EROS stereo viewer indicates that the limiting
resolution of the final image in an optical viewer using anamorphic and zoom lenses corresponds
to about 150 lines per millimeter in the center and 50 lines per millimeter at the edge of a field
about 8 millimeters in diameter, referred to the input material. This represents about 2 x 106
image elements, which does not differ greatly from the capacity of the electronic system.
1
29
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
W
0
Electronic System
Optical System
Feature
Available
Remarks
Available
Remarks
Sensing of parallax
Yes
Electronic scanning
No
Possible future use
of coherent light
Transformation capability
Zero order
Yes
Move scanning raster
Yes
Move input
First order
Yes
Distort scanning raster
Yes
Anamorphic lenses
Second order
Yes
Distort scanning raster
Yes
Tilted plane
Higher orders
Yes
Distort scanning raster
No
-
Gray scale processing
Negative-positive polarity
Yes
Invert video signal
No
Reflection method
inversion
not suitable for
transparencies
Contrast
Yes
Change video gain
Yes
Limited, change coherence
of illumination
Nonlinear characteristic
Yes
Change video
No
-
characteristic
Edge enhancement
Yes
High pass filter
Yes
Spatial filter
Shadow suppression
Yes
Logarithmic amplifier and
No
-
high pass filter
- Noise suppression
Yes
Low pass filter
Yes
Spatial filter
High resolution
Static imagery
Yes
1.6 x 106 image
Yes
2 x 106 image elements
elements
Changing imagery
No
Limited by bandwidth
Yes
2 x 106 image elements
of image transducers
System output
Direct viewing
Yes
Long persistence crt
Yes
Aerial image or screen
Hard copy output
Yes
CRT film recorder,
Yes
Film
high resolution
Remote viewing-printing
Yes
Remote crt
No
-
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
M r
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
3. GENERAL DESCRIPTION OF MULTIPLE IMAGE
INTEGRATION VIEWER-PRINTER
This section describes the present design of the multiple image integration viewer-printer,
with regard to producing this unit as deliverable hardware. Certain detailed design remains to
be completed in certain areas, and will be finished during the development effort.
The general configuration of the multiple image integration viewer-printer is shown in
Fig. 3-1, and a block diagram of the unit is shown in Fig. 3-2. The unit is housed in a console
78.5 inches in length, 69 inches in height, and 47 inches in depth. A viewing table projects from
the front of the console, and is 38 inches above the floor. Casters are attached to the console for
ease of movement, and access panels are provided to facilitate equipment adjustment and main-
tenance. The three separate input channels within the console each have a hinged control panel
located on the front of the console. The input platens for each channel are located on the viewing
table. Display units provided are the viewing monitor and the 5-inch cathode-ray printout tube.
A 4- by 5-inch camera can be attached to the face of the printout tube to provide a hard copy
record.
The control panel of the multiple image integration viewer-printer is shown in Fig. 3-3. The
function of each control is as follows:
? FILM DRIVE-drives the film forward, backward, left, or right
? ZOOM-selects magnifications of 0.925x, 1.66x, or 3x
? GAIN-varies the intensity of the video signal
? FLICKER-selects the flicker frequency, and switches the video on and off
? GAMMA-adjusts the shape of the video gain curve
? CORRELATION WEIGHTING-selects the square root, linear, or log function
? POSITIVE/NEGATIVE-selects the positive or negative video signal
? X in X-linear x-scale adjustment
? Y in X-y-skew adjustment
? XZ in X-oblique scale adjustment
? XY in X-keystone adjustment
? X in Y-x-skew adjustment
? Y in Y-linear y-scale adjustment
? XZ in Y-oblique scale adjustment
? XY in Y-keystone adjustment
The configuration of the three input platens is shown in Fig. 3-4. The three input platens
are located on the glass tabletop (1/4-inch-thick, " Parallel-O-plate" glass), which is illuminated
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
w
N
Viewing
aperture
r -
I---------- I
I
i
Magnetic ; 4== 1==
shielding
I ;~ I i I
~
'L_
I _
I i
1
69 in.
inch I
camera, I ,+t
31/2 inches, ; r
f/6.8 ~
h~\
?
}
5-inch printout
tube
.L_ --,
Fig. 3-1 - General configuration of multiple image integration viewer-printer
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Ml Ml M Mr
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
High
resolution
oS~ Video Video printout
processor combine
~Sp Video Film
processor
S1 Video
processor
Pre- Pre- Pre-
amp amp am Flicker Viewing
control monitor
~P--j Image dissector
( I~ILLLI ,
Image dissector
12
Video
processor
Master
oscillator
Video
processor
Pre
amp
Video
processor
Manual
viewpoint
control
Slow
scan
generato
Lissajous
scan
generator
Correlator Distortio
analyzer
Correlation
sequence switch
Fig. 3-2 - Block diagram of multiple image integration viewer -printer
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
LINEAR
SO. RT.0 0 LLOG
Q ED
0 POSITIVE/NEGATIVE 0
Fig. 3-3 - Control panel
CORRELATION
WEIGHTING
from below to allow viewing and positioning. Since the distance from the emulsion plane to the
lens must be within 0.002 inch for an f/2 system, close tolerance control of the platen will be ac-
complished by supporting the glass tabletop on an air inflated rubber sheet, which will sup-
port the glass plate evenly over the entire area. The height of the rubber sheet will be controlled
by three self-aligning exhaust valves which can be adjusted to provide a definite positioning of the
sheet, and consequently, the glass plate.
The platen can accommodate 9- by 9-inch and 70-millimeter film chips, which are placed
between two pieces of glass. The platen is translated in both the x and y directions by the joystick
control, through the use of ball screws and a motor-clutch arrangement. Disengagement of the
clutches allows manual positioning of the platen.
The section of film to be scanned is placed on the index registration mark, and when the platen
is actuated, the desired area of film will move under the scanning tube. Manual positioning can be
accomplished by handcranking or by positioning the platen by hand. The platen rests on three air
bearings that are operative during platen positioning to provide smooth movement.
The optical system (Fig. 3- 5) performs the following three functions:
1. Images the light from the flying spot scanner onto the film
2. Collects the light passing through the film and sends it to a photomultiplier tube
3. Images the film onto an image dissector tube.
34
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
1
~
1
I
_
I I
M
+ ~
II IT1
71,
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
v
Fig. 3-5 - Diagram of optical system
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
Functions 1 and 2 are used in driving the correlation circuitry, while function 3 is used to
obtain the high resolution video information. Function 1 provides the light source for function 3
by scanning the image dissector synchronously with the flying spot scanner.
Function 1 is accomplished with a lens designed for a P-11 phosphor spectral output, and
corrected for the 2:1 conjugate ratio. The lens will be a 75-millimeter-focal-length, f/1.9,
Wollensak Oscillo-Raptar. It will perform adequately at full aperture, although it may be
stopped down to f/2.8 to improve the depth of focus.
The phototube pickup lens for function 2 will be located below the film so as to be out of
the high resolution portion of the optical path. It will be a simple planoconvex lens with a
1-inch diameter and 2-inch focal length. The pickup lens will image the 75-millimeter lens
aperture onto the front face of the photomultiplier tube. A beam splitting pellicle mirror will
be used to deviate this beam onto the phototube. This pellicle mirror is manufactured by the
National Photocolor Company, and is used to avoid having a tilted, thick glass plate in the high
resolution system. The pellicle mirror also avoids ghost images from the back surface.
The high resolution system uses another 3-inch Oscillo-Raptar lens. This system would
ideally move the lens through a range of distances to give a continuous zoom capability, but this
is difficult to accomplish because of the need to dynamically maintain focus. For the purpose of
this equipment, it will be adequate to provide three discrete magnifications and to zoom the
raster size on the image dissector to fill in the gaps.
The lens magnifications chosen are 0.925x, 1.66x, and 3x. The lens performance, referred
to the film plane, will be better than 100 lines per millimeter at all points in the field, at all
magnifications. At the highest (3x) magnification the resolution will be better than 150 lines
per millimeter.
The correlation subsystem consists of the following components:
1. Lissajous scan generator
2. Photomultiplier tubes and video amplifiers
3. Video correlator and distortion analyzer
4. Memory unit
5. Modulator.
3.4.1 Lissajous Scan Generator
To detect image parallax, a scanning pattern must be used 'in which motion of the scanning
spot has components in both the x and y directions. The crossed diagonal, or Lissajous, pattern
formed by two triangular waveforms of slightly different frequencies has been found to be ideal
for this purpose. Each point on the image is scanned successively in four orthogonal directions,
enabling image parallax in any direction to be computed.
A scanning pattern using 500 lines across the diagonal will be used for this equipment, with
a line frequency of 15 kc/sec and a frame rate of 30 cps. A diagram of the Lissajous scan genera-
tor is shown in Fig. 3-6. The x and y frequencies are derived through binary dividers from a
common master oscillator, which ensures a fixed phase relationship between the frequencies.
This is necessary in generating a stable pattern. In addition to the x and y deflection signals,
this unit also provides the x and y reference signals that are required in the distortion analyzer.
The 0-and 90-degree reference signals at 15 kc/sec are produced by further dividing the normal
and inverted30-kc/sec signals at the output of the main divider by a factor of two. This ensures
an accurate phase difference.
38
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
15
me/sec
_ 90?
Inverter _ 2 ~---Ox
15-kc/sec
I Ooutput
2 x
30 kc/sec 0?
30 kc/sec Inverter
Fig. 3-6 - Lissajous scan generator
90?
-Oy
15-kc/sec
output
-Oy
0?
3.4.2 Photomultiplier Tubes and Video Amplifiers
The photomultiplier tubes pick up the fluctuating light intensity produced by density variations
in the film as it is scanned by the flying spot, and convert it into a varying electric current. This
current is then amplified by the video preamplifier and applied to the video processor. The video
processor has provisions for polarity change to accommodate both positive and negative input
material, gain control to account for differences in image contrast, and gamma control, so that the
shape of the video input and output characteristics can be changed to suit the input material.
3.4.3 Video Correlator and Distortion Analyzer
The video signals are correlated by separating the video into three frequency bands, each
channel covering about one octave. One channel of the video correlator and distortion analyzer
is shown in Fig. 3-7. After band filtering, the left and right inputs are phase shifted by
+45 degrees and -45 degrees, and hard limited so that only the zero crossings are preserved.
The two signals are then applied to an exclusive OR circuit. The output of this unit, when the
inputs are identical, consists of a symmetrical square wave. When either positive or negative
phase shifts are present between the video inputs, indicating parallax between the images, the
symmetry of the square wave is shifted to either positive or negative.
The output of the video correlator is applied to the analyzer modules. Parallax in the x and
y directions is detected by multiplying the video correlator output by the x and y reference sig-
nals, delayed by suitable amounts to equalize the delay of the video processor and correlator.
The x parallax and y parallax signals are further multiplied by the x and y 90-degree refer-
ence signals to detect the first-order distortion errors, i.e., x scale, x skew, y scale, and y skew.
The correlator and analyzer modules are mounted on printed circuit boards and extensive
use is made of solid state integrated circuit components.
3.4.4 Memory Unit
The memory unit (Fig. 3-8) stores the error signals from the distortion analyzer in the
form of the voltage to which a potentiometer is set. Six potentiometers per channel are required
to store thezero- and first-order error signals. The potentiometers are servo driven so that
the voltage at the potentiometer wiper equalizes that of the error input. The correlation sequence
switch initiates operation of each channel, in turn, by activating the servoamplifier. When the
potentiometer has been driven to equalize the error voltage, the null at the output of the differen-
tial amplifier is detected; when all channels have been nulled, the sequence control switches to
the next channel.
39
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
a
0
x reference, 90 degrees
x deflection reference
Left video input Bandpass Zero level
filter discriminator
I Video correlator
Right video input Bandpass Zero level
filter discriminator
y deflection reference
0
_y reference, 90 degrees
Fig. 3-7 - One channel of video correlator and distortion analyzer
Registration
error
signals
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
Null
detector
f
1`
From other
differential
amplifiers
Servo- m nth- moo,.,.,-i;f;o.a
Inputs from
distortion analyzer
To other servo u
Fig. 3-8 - Block diagram of memory unit
y shift
x scale
y scale
x skew
y skew
To modulator
circuits
3.4.5 Modulator
The function of the modulator is to distort the scanning pattern applied to the cathode-ray
tubes and image dissectors in accordance with the error signals stored in the memory unit. The
modulator unit shown in Fig. 3-9 consists essentially of several linear balanced modulators and
summing circuits.
The x and y scanning waveform inputs are controlled by the positive or negative error
input voltages, and are added to the main x and y scanning waveforms. For example, to com-
pensate for a positive x scale error, a certain proportion of the x scan waveform is added to
the basic x deflection, resulting in an increase in the x scanning amplitude. Similarly, an
x skew error is compensated for by adding a certain proportion of the y scan waveform to the
basic x deflection.
The modulator units are mounted on printed circuit boards and make use of integrated solid
state circuitry.
3.5.1 Video Processor
The output of the image dissector is fed to a preamplifier and polarity selector that permits
selection of either positive or negative display. The output of the polarity selector is fed to the
video processor. Fig. 3-10 is a block diagram of one of the three image dissector channels of
the multiple image integration viewer-printer.
The camera preamplifier is essentially an impedance transforming stage whose output is at
a level that enables it to drive a low impedance line terminated at the video processing unit.
amplifier
sc le
x0 skew
0y skew
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
y2 ref in
y scan
in
y scan
0
o y scan
to crt
o x scan
to crt
x scan a-
in FU
Y Y xy
scale skew in
Y
x x x2 ref xY
skew scale in in
x
First-order transformation
error signals in
0 0
y2 y2
in in
y x 111
x2 x x y
in in
y x shift
Second-order transformation
error signals in
Fig. 3 -9 - Block diagram of modulator circuits
Zero-order
error signals
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 M
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
Image dissector
? Focus coil
? Deflection coil
Deflection
driver
Scan
correction
Clamp
driver
Sync
generator
To flying
spot tube
Camera
video
preamplifie
Video
amplifier
Power
supply
Variable
gamma
correction
Aperture
correction
Fig. 3-10 - Block diagram of a single image dissector channel (sync
generator is common to all three channels)
W
Blanking
F-Lnrti
To switches Edriver
and display
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
The first stage of the processor is a wideband video amplifier feedback pair which steps up
the video voltage to the level required by the gamma correction amplifier stages. The block
diagram of a variable gamma amplifier used in our breadboard tests is shown in Fig. 3-11. As
shown in Fig. 3-11, the video amplifier drives two gamma (nonlinear) amplifiers in parallel.
One amplifier produces a video signal which is the square of the input, the other delivers a video
signal output which is the square root of the input. To establish a constant operating point, a
keyed dc clamp is provided, which sets the operating point (base bias) for the two gamma ampli-
fiers during the blanking period, which is itself established by turning off the illumination on the
flying spot tube during the turnaround of the scan cycle.
>_____II
From video
amplifier
DC pulse
clamp
amplifier
Gamma
amplifier
Gamma
amplifier
Fig. 3-11 - Block diagram of variable gamma amplifier
ay, + by2
2
Gamma corrected
video to line driver
The nonlinear load for the gamma amplifiers is made up of three resistors in parallel, with
diodes in series with two of the resistances, as shown in Fig. 3-12. The diodes are biased so
that they close at different predetermined video signal levels, switching in the shunt resistance.
Since the gain of the amplifier is proportional to the load resistance, this shunting effect lowers
the load as the video signal increases, producing a close approximation to the desired nonlinear
square root characteristic. Fig. 3-13 shows the transfer function characteristics of the non-
linear amplifier; for square law amplification the diode polarities are reversed.
44
Fig. 3-12 - One form of nonlinear amplifier
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
1
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
Fig. 3-13 - Transfer characteristic of nonlinear amplifier
The outputs of the two gamma amplifiers are fed to two inputs of a mixer, essentially two
differential amplifiers. The signals are introduced into the emitters of each pair. One collector
of each pair is tied together to a common load. A variable dc control bias applied to the base of
a transistor in each of the differential amplifier pairs controls the amount of signal from each
source. The output of the mixer has an approximately constant amplitude
aYa + = k
2 2
where a and b are the amplitudes of the two gamma corrected signals as set by the gamma con-
trol bias, and k is a constant. It can be seen that at one end of the gamma control the output is
yl, and at the other end it is 72, In the middle of the gamma control, the sum of the two signals
closely approximates a linear signal.
The signal from the gamma correcting stages is fed to an aperture corrector. As shown in
the general discussion of the MTF of the system (Section 2.3), the limit on resolution is in part
imposed by the finite size of the scanning aperture. The falloff in MTF caused by aperture
losses is phaseless, and is similar in character to the falloff in optical systems. It is possible
to partially compensate for this effect by generating a rising, phaseless response characteris-
tic as the video frequency increases in the video amplifier.
The amount of correction that can be introduced is limited by the increase in noise that can
be tolerated. In general, it is safe to assume than an approximate 6 db boost can be used at the
higher frequencies, resulting in an apparent increase in the sharpness of the reproduced picture.
In our MTF analysis, the effect of the use of aperture correction was not included.
After being fed through the aperture correction circuitry, the signal is fed to the blanking
insertion amplifier. The purpose of this stage is to insert a negative (black) pulse to blank the
printout cathode-ray tube during the retrace of the scanning beam. Control of the amplitude
of the blanking pulse is used to set the operating point on the monitor tubes.
Fig. 3-14 shows the results obtained with a step voltage signal applied to the input of the
gamma amplifier, and the gamma adjusted for each of the three conditions described.
Fig. 3-15a shows the waveform of a video signal with the positive peaks amplified more
than the black, in essentially a square-law relationship. Fig. 3-15b results from taking the
square root of the video signal with the black, negative peaks expanded.
The effect, subjectively, of variable gamma control is that of illuminating the shadows for
the rooting case, enhancing detail that previously was lost in more or less solid black.
45
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 CIA-RDP78BO477OA000200010048-8
(a) Signal A
(b) Signals (A + B)/2
(c) Signal B
Fig. 3-14 - Results of step voltage signal applied to input of gamma
amplifier (0.5 ms/cm, 2 volts/cm)
46
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
1
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
(a) Square-law correction,
white peaks expanded
(b) Square root correction,
black peaks expanded
Fig. 3-15 - Video signals
1
3.5.2 Experimental Gamma Amplifier
The schematic of the experimental breadboard gamma amplifier is shown in Fig. 3-16.
Two of these breadboards were constructed and used with the ARES for evaluation.
Tracing the cirucit from the video input, which requires approximately 0.1 volt peak to peak
of video signal, we see that the first two stages, Qi and Q2, comprise a feedback pair, with the
gain adjustable by means of a resistance in the feedback loop. The output of this stage is fed to
Q3 and Q10 in parallel, which are used to invert the signal b fore it is fed to the gamma ampli-
fier driver, Q5, in one channel, and Q11 in the other. The base of each of these stages is clamped,
by a pulse derived from the blanking pulse on the flying spot tube, to the bias set by the poten-
tiometer marked "black level adjust."
The nonlinear load for the gamma drivers is made up of three resistors in parallel. Two
of the three resistors have diodes in series, biased such that the resistance is switched in as
the video signal amplitudes overcome the bias voltage.
The outputs of the nonlinear amplifiers are coupled through emitter followers Q5 and Q11
to the adder stages, Q6 and Q12 in one channel, and Q7 and Q8 in the other. Each of these pairs
is essentially a differential amplifier, with Q6 and Q8 sharing a common load. The two gamma
corrected signals are applied to the common emitter load of each of the pairs. A do bias con-
trolled by the potentiometer marked "gamma" is applied to one base in each differential pair,
so that as it is increased the amplitude of the signal from one pair increases, while the ampli-
tude from the other decreases across the common load resistance. At one end of the gamma
control setting, the video signal is solely from one channel; at the opposite end, from the other
channel; and in the middle of the potentiometer setting, the two video signals are added in equal
proportions.
The line driver consists of Q13, Q14, and Q15. Output gain is controlled by means of a con-
trol in the emitter of Q13. Emitter Qi5, in series with Q16, is the driver with a low source imped-
ance, and permits the driving of several low impedance lines, as shown.
3.5.3 Video Combiner
The video signals from each of the three image dissector and video processor chains are
fed to the video combiner. At this point, combinations of the three signals are made for pur-
poses of registration in several modes: (1) superimposition, (2) flicker, and (3) individual
image viewing.
The mode is selected by means of a switch on the control panel, and the result is viewed
on the viewing monitor for the purpose of checking the registration before initiating the printout
cycle.
41
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
Electronic Superimposition
Several modes of combining the images from each of the three film input stations will be
provided. The simplest method of combining images electronically is by amplitude addition.
Thus, the signal from station A is added to the signal from station B, and, if desired, the sig-
nal from station C is added to the first combination. One necessary technical consideration
when planning the use of this method is the dynamic range of the video channel. When viewing
one channel, the signal amplitude normally occupies the entire range (from black to white) that
the video circuitry and the display cathode-ray tube can handle without saturation. It is there-
fore necessary that when signals are added together, the sum of their amplitudes be no greater
than the amplitude of each video signal viewed separately. This requires that as signal B is
added to signal A, the peak amplitude of signal A be reduced so that A + B = 1. Or, if B = A,
A and B must both be reduced to one-half the amplitude that would be used if either were to
be viewed independently. One signal can be subtracted from the other by reversing its
polarity.
The overall effect of this type of addition is that each of the two images appears to have a
lower brightness and contrast than when viewed independently. However, the capability of super-
imposition for purposes of registering several input sources is sufficiently valuable to warrant
this sacrifice.
Several methods are available for automatically limiting the amplitude of the combined
images to a value that prevents system overload. We will use a matrix that allows variable
gain in each channel, while holding the output amplitude constant for all amplitudes of input.
Flicker
Another method of superimposition is the addition of the several signals in time. Each sig-
nal is gated independently into the display system in sequence. If it were possible to do this at
a cyclic rate exceeding the eye's flicker sensitivity, the results obtained would be similar to the
effects obtained by the amplitude additive method described above. The time additive technique
is based on a law formulated by Fox Talbot in 1834:
"If a point of the retina is excited by a light which undergoes regular and periodic
variations, and which has the duration of its period sufficiently short, it produces
a continuous impression equal to that produced if the light emitted during each such
period were distributed equally throughout the duration of the period."*
In the design of television systems, flicker is avoided by presenting successive pictures to
the eye at a rate well above that at which flicker is noticeable. However, if two images that are
similar except for several small details are gated on in time sequence, and a balance is achieved
for the larger tonal areas, the resulting image will appear to be continuous except for those small
details that exist in only one image. These points will be presented at half the normal periodicity
and will flicker at a rate that makes them very noticeable to the observer. By means of this
method, successive photographs of the same area over a period of time can be quickly compared
for the purpose of detecting new targets or other changes in an area of interest.
To make fullest use of this technique, several conditions must be met in the design of the
electronic display system. These conditions relate to the geometry of the superimposed pictures,
the luminance values of similar areas in each image, and the repetition rate at which each image
is presented to the observer.
*J. W. T. Walsh, "Photometry" 2nd ed., Constable and Co. Ltd, London, 1953, p. 67.
1
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
Approved For Release 2004/07/29 CIA-RDP78B04770A000200010048->
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
1
1
These conditions are interrelated. The repetition rate at which each image is flickered or
gated into the display system is, for optimum detection, dependent on the luminance of the display.
For low light levels, the critical frequency of flicker is much less than for displays of high lumi-
nance. The color of the display and the angle of view are also factors. The simplest case com-
prises periods of stimulation by a constant brightness alternating with equal periods of complete
darkness. The lowest frequency at which no flicker is seen is defined as the critical frequency.
Although this point varies from one observer to the next, at a luminance of approximately 1.0 foot-
lambert, it is usually found to be between 12 and 15 flashes per second. A difference in luminance
of 11/2 percent between the two images is detectable at these frequencies.
From the above discussion it may be readily seen that the luminance of the two images to be
analyzed must be adjusted such that the luminance of similar areas on the display screen differs
by less than 11/2 percent. The requirement for precise geometrical registration is necessitated
by the fact that misregistered points will flicker.
Because of the characteristics of the electronic display system, it is necessary to select
flicker rates that permit the display of an integral frame, or a multiple of integral frames, dur-
ing the "on" period. Since the television frame rate is 30 cycles per second (60 fields interlaced
2 to 1) we can choose an "on" period of any integral multiple of 1/30 second. With one cycle rep-
resenting the viewing of the two time- superimposed images, this limits us to flicker rates of 15,
7.5, 3.75, and 3 cycles, etc.
It is not recommended that flicker rates be chosen to include an odd number of fields per
cycle, since interline flicker between the interlaced fields will be annoying during observation of
fine detail and may have somewhat the same effect as geometrical misregistration of the images.
The flicker display mode is implemented by an electronic switch. Switching occurs during
the period used for vertical retrace on the display screen, and is initiated by appropriate logic
circuitry that is triggered by the vertical drive pulse. Selection of flicker rate will be on an
individual channel basis that permits the operator to select the time during which each channel
will be "on" in any integral number of frame periods. For example, it will be possible to switch
channel A on for two frame periods and channel B on for one, by setting the channel A switch to 2
and the channel B flicker control switch to 1.
The flicker rate is variable in submultiples of the frame rate and is controlled by an elec-
tronic counter locked to the Lissajous scan converter. The output of the electronic counter
operates an electronic switch in each channel. Fig. 3-17 shows the switching of two signals. For
the purposes of illustration, one input signal is a 1-megacycle sine wave, the other a noise signal.
Fig. 3-17 - Switching for flicker presentation (0.2 ms/cm, 1 volt/cm)
The rate of the switching for flicker presentation is controlled by the counter which is locked
to the frame time of the scan generator. The shortest duration of switching is one frame time.
The duration can be selected in multiples of the frame time. This allows the duration selected
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
to be 1/30, 1/15, 1/8, 1/4, or 1/2 second. Thus, when two channels are being viewed, the flicker
rate can be varied from 15 cps down to 1 cps. It is also possible to select different periods for
each channel, so that, for example, one channel is viewed for two frames, and the other for one.
In addition to the variable flicker rate, permanent on and off positions are provided on each con-
trol. By this means, one channel at a time may be flickered on on top of a continuous display by
the remaining channels. This method of control is very flexible and avoids the display of incom-
plete frames.
The expected performance of the complete multiple image integration viewer-printer system
is summarized below.
3.6.1 Input Material
The viewer will accept three film chips of any size up to 9 1/2 by 9 1/2 inches-positive, negative,
or mixed.
3.6.2 Hard Copy Output
The output will consist of a 3- by 3-inch image recorded on standard 4- by 5-inch cut film
or Polaroid material. A regular 4- by 5-inch camera back will be used.
3.6.3 Viewing Outputs
In addition to the film output, two viewing outputs will be provided for monitoring purposes:
1. A 10-inch-diagonal, short persistence crt scanned with the 500-line crossed diagonal
raster, at 30 frames per second, to check operation of the automatic registration circuitry
and to facilitate superimposition and gray scale adjustments
2. A 10-inch-diagonal, long persistence crt scanned with the 4,000-line, high resolution
raster, at about 1 frame per second, to check operation of the slow scan readout and out-
put image quality.
3.6.4 Magnification Ratios
Three magnification ratios will be provided as detailed in Table 3- 1. In addition, the size of
each input will be electronically adjustable over a range of 2:1, to accommodate variations in im-
age scale and geometry.
Table 3-1 - Input Magnification Ratios
Magnification
Magnification
Image Area
Onto Film
Onto Viewing CRT's
Viewed at Input, inches
12x
28x
0.25 by 0.25
6.7x
15.5x
0.45 by 0.45
3.7x
8.6x
0.80 by 0.80
Modulation transfer functions of the complete system have been calculated and are shown in
Figs. 3-18 and 3-19. Fig. 3-18 shows the individual MTF's of the basic components of the system,
52
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
all referred to the face of the image dissector tube. Fig. 3-19 shows the overall theoretical MTF
of the complete system at the three magnification ratios of Table 3-1, referred to the input.
3.6.6 Signal to Noise Ratio
The signal to noise ratio of the system will be at least 100:1 (40 db). The signal to noise ratio
of the output image will theoretically be increased by a factor of 1.73 over that of a single input
when three inputs containing random additive noise with identical signal to noise ratios are
integrated.
Camera lens
Image dissector,
1-i
1- b
h
t
nc
y
ras
er
CRT, 1.5-mil spot,
3- by 3-inch raster
Image dissector lens
,
maximum magnification
Combined response,
maximum magnification
Spatial Frequency at Image Dissector Tube Face,
cycles per millimeter
Fig. 3-18 - Modulation transfer functions referred to image dissector
53
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
0.25- by 0.25-inch --,
image area
0.45- by 0.45-inch
image area
1,000
Spatial Frequency Referred to Input,
cycles per millimeter
Fig. 3-19 - Theoretical modulation transfer function of complete system
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
1
1
1
1. American Society of Photogrammetry, "Manual of Photogrammetry," George Banta Publishing
Company, Menasha, Wisconsin, 1952, pp. 319-324, 336-345.
2. Perkin-Elmer Corporation, The Practical Application of Modulation Transfer Functions,
Phot. Sci. Eng., 9(4):235-264 (July-Aug 1965).
Scott, F., Scott, M., and hack, R. V., The Use of Edge Gradients in Determining Modulation-
Transfer Functions, Phot. Sci. Eng., 7(6):345-349 (Nov-Dec 1963).
9. Swing, R. E. and Shin, M. C. H., The Determination of Modulation-Transfer Characteristics
of Photographic Emulsions in a Coherent Optical System, Phot. Sci. Eng., 7(6):350-360
(Nov-Dec 1963).
10. Wentworth, J. W., Color Television Engineering, RCA Victor Division, Radio Corporation of
America, Apr 1962, pp. 7-8-7-13.
Kohler, R. J., and Howell, H. K., Photographic Image Enhancement by Superimposition of
Multiple Images, Phot. Sci. Eng., 7(4):241-245 (July-Aug 1963).
13. Higgins, G. C., Lamberts, R. L., and Wolfe, R. N., Measurement and Analysis of the Dis-
tribution of Energy in Optical Images, J. Opt. Soc. Am., 48(7):487-490 (July 1958).
14. Lambert, R. L., Relationship Between the Sine-Wave Response and the Distribution of Energy
in the Optical Image of a Line, J. Opt. Soc. Am., 48(7):490-495 (July 1958).
15. Lamberts, R. L., Application of Sine-Wave Techniques to Image-Forming Systems, J. SMPTE,
71(9):635-640 (Sept 1962).
55
Approved For Release 2004/07/29 : CIA-RDP78B04770A000200010048-8
STAT
STAT
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
Appendix
It is supposed that a number, N, of photographs containing the same signal, y, are to be
superimposed so that the output, S, is
N
S = ai(aiy + ni)
i=1
where ai represents the signal amplitude and ni represents the noise signal on the ith input. The
problem is to determine the coefficients, ai, to maximize the fidelity of the representation, S.
In optical signals, the dc noise level represents an overall gray scale and can result in a re-
duction of contrast. The detail, however, is not affected by the dc levels; it is degraded only by
variations in the noise signal. The method employed here is to determine the ai's so that the re-
sultant noise variance is minimized for a given output signal strength.
1
1
1
N 2 2
Oi9i
~'2 a
i=1
51
N
S aly + ni ai
where Qi = aiai
The output signal strength is then fixed by constraining the /3i's so that
N
ai=k
i=1
Let 92 be the variance of the ith noise signal. By noting that the variance of the noise signal,
aini/ai, is (3jc4/ai, and using the fact that the variances of the different inputs add (because of the
independence of the noise samples), we can write the variance of the output as
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
Using the Lagrange multiplier, A, to take care of the constraint, we obtain
N 22
2 = E aiai
or a -~ ai-k
i=1 i=1
Minimizing with respect to a particular coefficient, say /3j, we have
2
2 2a;oj
bQ = a -X 600
y=
i
Since 6(3j is arbitrary, the function in the parentheses must vanish, and the result is
~
aX
Rl 202
J
The value of A can now be found from the constraint equation
N N 2
[~1 ai
2
L ai k
Q
i
i=1 i=1
2k
M
It should be noted that M can be determined from the inputs alone. The solution then is
2
k ai
Pi =Mx 2
of
k ai
ai=Mx 2
of
From this last expression it is seen that the variance of the output is minimized for a given output
signal amplitude when the coefficients ai are proportional to the input signal strengths, ai, and in-
versely proportional to the noise variances, Qi. The constant k/M determines the intensity of the
output and need not be computed in practice.
1
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
In the particular case of equal input signal amplitudes, ai = a, and equal noise variances,
9i 9 , the result becomes
ai = k x ; = constant
Q
The coefficients ai and ai then become superfluous and a single input can be represented by
y+ni
By simply averaging the N inputs, we obtain the output
N
S=y+N Ni
f=1
Since the noise samples are independent, the variance of the sum is the sum of the variances.
The output variance, therefore, is
N
2 1 2 0,2
Qo= L Q =N
i=1
and the standard deviation from average of the output noise signal is
Qo = NNi2
1
1
Hence, in the case of equal input signal strengths and equal noise variances, the noise attenuation
factor is the square root of the number of inputs.
Having found the ai's, we can find the output signal to noise ratio in terms of the input signal
amplitudes, ai, and the input noise variances, vi. The output is
f2
Sk al
Y 2
of a1
M of
The signal power is the integral of the square of the signal part of the output
S N z
kz of
Ps = f( -2 Y(, 71) d~ dry
i
N 2
2
k2 ai
i=1 T
59
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
where T is a constant given by
T = f5Y2(x, y) dx dy
The noise power is the integral of the square of the noise part of the output
N 2
k2 ~1ai
Pn= %Q, 77)
oi
i=1
When the dc components of noise are neglected, the above expression
that all cross product terms in
samples and hence integrate to
Pn =
can be simplified by noting
the square sum represent cross correlations of independent noise
zero. Therefore,
ai
N! (x, y) dx dy
(455
he power ratio of signal to noise then is
T
Ps 2
=T z
Pn of
E
In the case of equal input signal amplitudes and equal noise variances, the above expression
becomes
PS = NT 2 2
n
The amplitude signal to noise ratio is the square root of Ps/Pn
N 1/2
2
A R _ As _ T1"2 \L\ ai
N An 01
i=1
where the subscript N indicates that it is the ratio based on N inputs. If an additional input having
signal amplitude aN + 1 and noise variance vN + 1 is added, the new signal to noise amplitude
ratio is given by
2
aN+ 1
R N 1 =RN+ T
U2
N+ 1
It is apparent that no input can degrade the output. The worst possible case is encountered
for an input having zero signal amplitude, in which case it is added in with zero weight and
N 2
t
1
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
 Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8
RN+1=RN
1
1
1
1
In the case of equal input signal amplitudes and equal noise variances
(NT) 1/2
RN = Cr
Approved For Release 2004/07/29 : CIA-RDP78BO477OA000200010048-8