DEVELOPMENT OF AN IMAGE-PROCESSING RATIONALE

• Document Type: CREST
• Collection: NGA Records (Formerly NIMA)
• Document Number (FOIA) /ESDN (CREST): CIA-RDP78B05171A000800070025-8
• Release Decision: RIPPUB
• Original Classification: S
• Document Page Count: 10
• Document Creation Date: December 28, 2016
• Document Release Date: November 17, 2003
• Sequence Number: 25
• Publication Date: June 20, 1969
• Content Type: MF

Approved For Release 2 2'PP9 : CIA-RDP78B05171A000800070025-8
IPIC/TSSG/QED-1676-69
20 June 1969
Chief, Development & Engineering Division
Development of an Image-Processing Rationale
fulfillment of the requirement to examine the field of
ocessing and its potential applications to NPIC operations, I
ed the following organizations to collect data relative to present
processing capabilities and future i:ossibilitiea:
5X1
Jet Propulsion Laboratories, Cal. Tech.
Pasadena, California (15 May 1969)
2. In addition to these visits, conversations on the same subjects
'era held during the Los Angeles conference of the Society of Photographic
cientista & Engineers (12-14 May) with the following individuals:
3. Definition
a. In the general case, image-processing is defined to mean any
operation that transfox a given image into another image in a quantitative
wanner.
b. As used hexin the term will be used to indie*te optical and/or
digital image transformation.
Declass Review by
NIMA/DOD
X1
Approved For Release 2003/12( -CJA-RDP 00070025-8
11''~EG ET downgrading and
Aesrassl lotion
 Rationale SECRET
Si1RJECT ; Development of an Image-Processing
Approved For Release 2003112/19 : CIA-RDP7>,05171A000800070025-8
25X1
program are.
4. 4ummsr of the Visit to JPL
Space Sciences Div.,
a. Introduction - ing resent
PL's im?ge_praces bac round, p
spent several hours explaining
hardware/software capabilities, objectives, and. rationale.
works inly with system hardware while his colleague, ]3r.
handles software development. processing
JPL has developed hardware and techniques for Their capacity is
Ranger;, mariner, Apollo, and other An spaced ai imagery has also been
not confined. to space imagery, digital as oppased to a
performed. The primary reasons for employing
coherent optical processor were the advanced state of the digital art
when the program began, the greater flexibility available digitally, and
the absence of the rapid processing time requirement.
b. Qbectives - The two main technical objectives of the JPL
(}.} To improve images by removing photometric and geometric
distortions, noise, and correcting for camera system
MT ;
(2) To provide high quality information extractio options;
e.g., contouring, edge enhancement, gray
tion.
c. flat Review '- and
expensive.
is comprehensive, Sophisticated, time coning,
(1) input/output
a. At present input/output options are limited to
emetic tape and a precision CRT; i.e., the Video Tape
b
ent
Film C?
nt o
er
converter and the Video ~
e
ptical 4,e a
25X1 Indus on
processor has been considerinputt not implemented. The
an d output opratio
hardeopy, and orbticsl devices can be
s equi t is
Images from film,
scanned, digitized and entered in the computer e. Telemetry
data can enter directly from magnetic
phasis has been placedated when-ing
b. Considerable em
a high signal-to-noise ratio. This is e definition The ever film is droduced*related to the resolution
signal-to-noise ratio.
attainable are
SECRET.
Approved For Release 2003/12/19 : CIA-RDP78BO5171A000800070025-8
 Development of an Vie-Processing Rationale SECRET
Approved For Release 2003/12/19 : CIA-RDP78p45171A000800070025-8
co JPL employs 1024+ picture elements (pixels) per unit
area scatted as opposed to 64 employed by[ and 256 by
In general the outer size is a major
I=tUkg factor. PL requires the increased speed and quality
batcause the unit area scanned (i.e., the largest area of an image
that can be scanned :zt ne time) at a film image resolution of
100 l/matt, is only 2.5 mm and the pictures produced must retain
suion definition. For a high quality system, using 6 bit
digitising and ; sampling rate o 20,000 samples/sec, 50 seconds
are refired to input the 2.5 mm unit picture area, From this
at W It it can be seen that time becomes another liziti
e ~sideration - remember the computer processing and output
time rests have still to be added in. Images of larger
areas are scanned in multiples of the unit area.
(2) The Computer
oys an 11 360/44 computer equipped with 4 tape
c drives, and other supporting components. Software
le imMe enhancement,, nonlinear transformation, sp
formations, and similar operations has been
2 lists potential application areas for digital
No average estimate of processing time can be
oration of each operation will vary 4th image
analysts judgment; however, times ranging from
uld not be extraordinary. The computer drives
devices and controls peripheral components. Processing
(3)
a proportional to computer capacity.
Display Console
This cosoponent of the JPL system is not yet operations:
oy an IM 2250 graphic display console with light ;.en.
known as the Interactive display console, it is
this component that permits man's judgment to enter the image-
pr'ocessl sequence; e.g., the selection of a correction operation
to be performed on a distorted image. Through the interactive
display unit algorithms :re constructed by trial and error.
The input image is digitized processed and the :,lgorithni recorded.
The input and output images can be fed to the console so that
before and after comparisons may be made.
(4) Discussion
research in image processing at JPL has until recently
been entirely sponsored by NASA. The National Institute of Health
is presently sponsoring a project with NASA consent.
indicated it would be possible to accept projects from other Government
25X1
25X1
Approved For Release 21S3O T: CIA-RDP78BO5171A000800070025-8
ITE
 p
`rov d 4oorfRRa gl1 fg~t154-MV"W805171A000800070025-8
agencies. 1Vor example, JPL is looking for support to
build a new high resolution scanner using an image director
tube and a CRT.
b. I viewed samples of image enhancement, ch>.nge
detection and amplification of small density differences.
The processed imagery was noticeably improved. No processing
t imagery similar to that viewed at NPIC has been done and
25X1
case, the lat er bas h e hi er resolution gained by
trading off with speed. The line-scan project may shed
he. Via
mitts would be required to determine whether the
electronic scanning used could retain adequate detail. In
ent configuration it is several orders of magnitude
he optical-mechanical scanner. Of
I
of the Mathematics Drall4ation Studies Department presented a de a ed
X1 briefing and. equipment demonstration which covered the present .And future
X sttue of im,:rge-processing at For convenience, their program was
divided in two parts. Part one, the
~ Iincludes a computer, an interactive display unit and the associated
25X1 '' acre. Part two, the includes
25X1 an automated coherent optical system which the image of diffraction
ttern may be viewed alternately, and a remote control console. The II
25X1 system was designed end developed by starting in 1961. In 5
submitted a proposal recommending e r system be applied to the
MC :?-2R rogram eventually awarded to Since that time
ss been refined and automated. The entire program has been
hed without external funding.
b. Program Objecgtivee - The program has two primary objectives:
(1) to develop a rapid automated signature analysis
capability with an adaptive learning network dependent,
in part, on human intuition and recognition abilities.
o develop applications in :addition to analysis of
ides on film.
SECRET
25X1
25X1
~ 25X1
25X1
25X1
25X1
25X1
25X1
Approved For Release 2003/12/19: CIA-RDP78BO5171A000800070025-8
 OLbfL I
SUBI; T : D lope-nt of an Image-Processing Rationale
Approved For Release 2003/12/19: CIA-RDP7Zp5171A000800070025-8
A berent optical system was chosen as an input
vivice because of its ability to handle or process large
System Review
(1) boat
aa
quantities of data rapidly; the input time -:nd Fourier
transformation are done at the speed of light. This approach
it not as versatile as the digital processing in terms of
the variety of transformations that may be accomplished (see
fig 2) and unfortunately, noise is still a major problem.
25X1 The illuminates film 0.6" in diameter with a laser beam.
The areas viewed may be selected at random by an operator
25X1
25X1
25X1
_,a_.. --.- ------ - --
ognized. in the future. To do this, the operator views
SECRET
by the computer. The diffraction pattern of the erea.2
t
on
e film is formed optically in a precisely known Puy
R,el rdr IV++ ...v --c-. -- --.-._- ._ -.
The output of the phototube is fed to the computer. Assuming,
for the moment, an algorithm exists for this particular area,
the computer will recognize the area (e.g., natural, man-made
fam) and print a description on the graphic display unit.
This entire sequence can occur in less than 10 seconds
depending on the complexity of the area examined and the
operator's experience. By comparison, Just the input
scanning phase of the JPL digital system requires stout
an hour.
(2) Computer (processor)
wry of 256K bytes and a lower speed large core storage
of one million bytes, of which the0 uses 400K. The
computer has six basic routines to handle the signature
analysis operations. No enhancement, change detection or
similar programs have been written at 0 since they are
already available from commercial and academic sources.
(3)
ut can take two forms - video (IBM
t
p
csu
display console) and printed (on computer paper).
r is displayed, only a matrix listing the images
.
nd their Identification. It is at this station
,,.,,.
.W Iowa"
in the system that the human interaction comes into play.
s
h
a
l
e u
.a
----
ed, on v
be operator in signa
w_ _ t%-L .26. -4'9'1 A?.
25X1
25X1
Approved For Release 2003/12/19 : CIA-RDP78B05171A000800070025-8
employs an I EM 360-65 computer with a fast
 piyt,Fefe
the picture and its diffraction pattern on the optical
system ground glass and attempts to determine a unique
correlation between the two using the basic computer
routines. With experience an operator can find such
carrel tions. ktlI this has been accomplished for
the natural, man-ma a agricultural categories and some
sub-categories; e.g.,, aircraft, forms, residential areas.
The algorithm building is iterative, intuitive and at
this point essential. It can be time consuming, but once
eluded, less than 1% wrong decisions are made.
Pecos.. can
; J Pi j A lE3i8BQ5171 A000800070025-8
system is operational. The 1I can be
or about I minas computer), rented or used
h tool on a contract basis. The IIsoftware
also be purchased. has been informed.
could-
a aM` that permits the recording of the portion of the
ire the operator looks at, in terms of image coordinates
and spatial frequency distribution, with the data collection
occurr' automatically. This approach could produce
eb.etive data concerning what the PI thinks is important
and how he goes about interpreting the image; a kind of
'ICOF" experiment within the image.
(2) The computer is interfaced with the optical system
of Visit to
;eived an update briefing on b- 0 " the approaches
(optical and electronic) to the -QTR problem from He is
aware that image-processing will be employed to eccomplah objectives beyond
their immediate program goals as the technology is developed and feels
is in a good position to participate in any such effort. x:s might
cted, the techniques and much of the film handling equipment developed
'byfor their present contract could be adapted to more general image
process tasks.
7. im of Visit to
built by including the input/output equipment for JPL. The
Marketing Manager, and (a systems
briefed me on a variety of information retrieval and analysis
items, of interest was their laser scanner. It is a breadboard item
present, and has been under development for several years.
25X1
25X1
25X1
25X1
25X1
25X1
25X1
Approved For Release ?9 : CIA-RDP78B05171A000800070025-8
 g_- -~ for NPIC application or tiQ WCXK : =s,,. ~.. -- -
high enough laser scanning techniques are the high signal-
primary reasons for pursuing available.
rate and the large amount of energy
m
Problems have been encountered in mused vithh atrotatingsprism controlling
t
25X1
25X1
stems a lined, as presen ly
s
i
e
y
ng
ss
e-of the image Proc
h the tasks listed in fig 2 rapidly enough to be
i
s
13 accompl
te exploitation on-line, or off-line. The systems are
.
pecial purpose at this time*
b. There is a consensus among the individuals mentioned herein
PI ima exproccessing
The technology necessary to "apt current
iaues to provide a highly automated, rapid,
mly the funding is J.acsszzg.
(2) Internal funding will not be adequate to develop the
d
.
rdware and software require
(3) A system Which sleets the criteria of rapid and versatile
operation will require -, variety of irkput/output options, a large
central computer, and a human interactive element .
25X1
(4) The time to start is now.
a
has excellent potential as
e The
ridge t e aubjeetive-0biective measure gap.
~l in,experi is designed to bridge-the-
of this idea.
i
on
ing a technical evaluat
y
Though the present
curved fired plane is curren
he beam. Only bar t9rgetshave been scanned to date.
..Ud* ^e a vear ago, it does not appear to be
to-noise ratio, rapid scan
ea er thth0ubzect will
better estimate of potential value will be made
scanner capabilities have been determined.
be attended on 1 July 1969 at Wright AFB.
8. Miscellar
q, C, elusions
e D
SECRET'
Qved, %TR,@Je Z i=ft* BO%171A000800070025-8
I I described a new type soda staL%! -
`?"u-using light sensitive P o ansitors in a linear array. Resolution values approaching 40 OYIMM are being tested now, 20 Cy/mm scanners have been ince developed. This approach holds great promise since scanningceiiesJofy100
ft//min are possible. A meeting Q
25X1
Approved For Release 2003/1 az. I-RDP78B05171A000800070025-8
 runtI
Develo
Approvpg~ F t of a a efIm / / 60 171A000800070025-8
ase
25X1
d, There are other image-processing systems that should be
ed before specific recommendations concerning their application in
evelopment of a t pr?hensive image processing program as opposed
to a
.
eries of small loosely related contracts.
e. The solid state scanner is worth a very close 1rr Tt
could result in a breakthrough solution to the long input time problem in
digital processing systems. It has many potential advantages over the
laser scanner.
10. B=e ion
July 1969
b?usst - September 1969
(2) Recommend preliminary programs as necessary.
(1) Complete the survey of firms listed in the egnclusi ns.
Prepare a comprehensive image-processing program con
NPXC objectives and the technological state-of-the-art.
25X1
T G/D
Distribution-
Original - ;addressee
2 - RPIC/TSSG/DED
25X1 NPIC/TSSG/DED (24 June 1969)
Approved For Release 2/19 CIA-RDP78BO5171A000800070025-8
 Approved For Release 2003/12/19: CIA-RDP7$&5171A000800070025-8
POLAROID
'HARD COPY' VIDEO MAG TAPE SCAN
CAMERA MONITOR VIDEO TAPE RATE CONVERTER
II
f!'1
I I
L.J
OPTICAL
MICROSCOPE
L_J INTERACTIVE
ELECTRON DISPLAY AND
MICROSCOPE CONTROL
CONSOLE
-- LINE PRINTER
CARD READER
TYPEWRITER
CARD PUNCH
INTERNAL
DISK.
DIGITAL
MAG
TAPE
w
INTERACTIVE
VARIABLE DISPLAY AND
MAGNIFICATION CONTROL
SCANNER CONSOLE
FILM AND
PHOTOPRODUCTS
IT. P. L.
IMAGE PROCESSING SYSTEM
Approved For Release 2003/12/19 : CIA-RDP78BO5171A000800070025-8
 Approved For Release 2003/12/19 ?: CIA-RDP70ZA5171A000800070025-8
Intensity manipulation
Geometric manipulation
0 t,
Spatial frequency operations
Analysis
Multi-picture `'
Intensity calibration of systems
Nonlinear lookups
Chromaticity calculations
Geometric calibration of systems
Reprojection
Overlay match of 2 pictures
Independent X and Y adjustments
Spatial high frequency boost
Spatial low frequency reduction
Fourier transform
Image light distribution
Pattern extraction
Convolution
Test targets .
Graphical displays
Insert windows in pictures
Insert good data for bad
Photometry
Film curve corrections
Grey scale alterations
Color shift, balance, alteration
Good geometry needed for stereo
Convert slant pixels to ground projection
Rubber sheet stretching
Aspect ratio corrections
Correct for detail losses in system
Minimize broad-brush shading
Remove effects of glare
Remove coherent noise
Analysis in spatial frequency plane
Star cluster analysis
Counting blood cells, automobile, stars, etc.
Analyzing shapes of objects
Filtering, correlation
Change detection
Stero information extraction
Addition Averaging, noise removal
Multiplication Spatial domain filtering
Division Normalizing
Approved For Release 2003/12/19 : CIA-RDP78B05171A000800070025-8