(Sanitized) PROPOSAL FOR TESTING OF DIGITAL SIGNAL-TO-NOISE OPTIMIZATION TECHNIQUES AT A COST OF(Sanitized)

• Document Type: CREST
• Collection: NGA Records (Formerly NIMA)
• Document Number (FOIA) /ESDN (CREST): CIA-RDP78B05703A000200020021-9
• Release Decision: RIPPUB
• Original Classification: S
• Document Page Count: 3
• Document Creation Date: December 28, 2016
• Document Release Date: December 22, 2003
• Sequence Number: 21
• Publication Date: April 9, 1970
• Content Type: MF

Approved For ReI se 2004/02/11 : CIA-RDP78BO5703A00920002002'1 ~~ - a' 7
IPIC/D-98/70
J APR 1970
MEMORANDUM FOR : Ass:i-stant Deputy Director for Intelligence
25X1
SUBJECT : II Proposal for Testing of Digital Signal-to--Noise
Optimization Techniques at a Cost of
1 -1
1. This memorandum requests approval for the commitment of funds
for a contract. The specific request is stated in Paragraph 9.
2. The size and distribution of the silver particles that comprise
a photographic image limit the fine detail that can be recorded on any
photographic emulsion. Images examined under high magnification have a
granular structure which gives rise to the term "graininess." When the
examination is performed by an instrument which scans in a raster mode,
like a TV, the grains cause random fluctuations in the output trace or
image which are not related to the object originally recorded; i.e.,
they produce an error signal. The more grain, the greater the random
signal fluctuat;i.ons and the greater the loss in fine detail. The situa-
tion is analogous to the white noise or specks seen on a TV receiver
with a poor antenna. Such a scanning operation is employed in microden-
sitometry and in digital- image manipulation in order to transfer the
image on the film into a useable input to the computer. In both cases
the transfer must occur with minimum error; in one situation, the stru.c-
ture of the image is being investigated, while in the other, the input
image is already degraded and. an improved representation is being sought.
2. Currently, we attempt to minimize errors due to grain noise by
a mathematical averaging of the data in the computer after scanning.
Although some improvement is achieved, it is not adequate for EPIC
operational applications. Better techniques are necessary and infor-
mation theory says they are possible. Historically, new methods for
noise reduction have required more data points (over-sampling) to achieve
limited success, thereby increasing processing time and computer storage
requirements. What is needed is a technique that will simultaneously
reduce the effects of noise, decrease the number of data points required,
25X1
25X1
and improve the system accuracy. Addressing this problem, the
eveloped the OCriterion, a mathematical computer al-
gorithm named after the originator, The first photographic
applications were to= microdensitometer traces being transmitted from
a West Coast to an East Coast computer for analysis. Typically, 135 data
points were required per trace. Applying the 1 Criterion to the trace
Declass Review by NIMA/DOD
pproved For Release 2004/02/11: CIA-RFj-76Q703A00020002 br,
~
25X1
25X1
25X1
25X1
25X1
 Approved For Release 2004/02/11 : CIA-RDP78BO5703AO09200020021-9
SUJi1C'_i_`: ru-po?al for Testing; of Di :i tal. S:i final.-to-No:i.se Opt in.`:.zatior
Techniques, at, a Coot of
data, only- 25 data points were necessary, and the accuracy of the trace
:i_mpr. oved. Simi:I.ar application of the Criterion to NPIC microden--
s:i.tometer traces would improve the (I uan i.lcation of the parameters used
in photo-optical image evaluation--thc only advance of this magnitude
developed in the last five years .
il. i'he project proposed is concerned with improving the accuracy of
information extracted from photographic images during the scanning and
digitizing operations employed in microdensitometry and digital image man-
ipulation (DIM). Extending the = Criterion concept to the I DIM
software would visibly improve the quality of' the image manipulated and,
at the same time, reduce by 50 to 60 percent the computer core storage
required, with a corresponding reduction in computer processing time. An
25X1
25X1
25X1
improvement in computer processing efficiency is especially important. One
square inch of operational NPIC imagery digitized. to only two gray levels
contains enough data points to fill the core storage of over sixteen Univac
),1.9)1. (32K core) computers.
5. This project is designed to be executed in :Four phases over
four months. Preliminary experiments would apply the = Criterion to 25X1
one-dimensional data, such as edge traces of "CORN" resolution targets.
Success in this application would lead to extension of the algorithm to
the two-dimensional case suitab:Le for use with the 0 DIM software. 25X1
Because the first DIM experiments on operational. material will be con-
ducted in NPIC during this period, operational tests of the = Criterion 25X1
would be integrated with that effort. Computer time required to compile
and. test the algorithm would be provided by 0 Software will be written 25X1
in Fortran IV suitable for use on the Scripps 360-4- computer. = will. 25X1
train NPIC personnel in the detailed use of the algorithm. The 0 25X1
Criterion would also be tested with DIM programs previously written by
I nd preliminary-, unclassified tests would be conducted to evaluate
their potential to NPIC applications. will deliver a comprehensive 25X1
technical final report detailing both practical aspects and the theo-
retical basis of the =Criterion. In addition, a one-day tutorial
session by =will be given early in the project to impart essential
expertise to NPIC personnel.
6. The theoretical foundation of the 0 Criterion is sound, and 25X1
the initial results are positive. The value of extending the technique
to NPIC imagery cannot be estimated quantitatively, though the potential
is very high. In,addition, it can be applied to other community digital
enhancement efforts as well. as to color imagery, to electronic signals
from MTF analyzers currently under development, and to the generation and
fabrication of more accurate filters for optical image manipulation systems.
The potential improvement is considered to be worth the low risk involved.
2
Approved For Release 2004/02/11 1 A 8BO5703A000200020021-9
 Approved For'Release 2004/02/11 : CIA-RDP78B05703A000200020021-9
=Provos l for Testing i vital Sigrial-to?-Noise Opt iim~ zat on
Techn:i_ques at a Cost of
The iinsol_ici.ted. proposal fr.orn =o:i'fer.; a unique opportunity to improve
the product of' a variety of key projects.
7. The = Criter-on is an =proprietary technique. Accord-
ingly, no other proposals were considered. The funds required were pro-
grammed as part of the Y)[-'(O R&D budget. The project has been closely
coordinated to insure that it will not conflict with the 0 DIM
project or the routine use of the APSD microdensitometer. Informal con-
tacts with AFSPPF, Westover. At'E, OED/DDS&`.C, TSD/SSP, NRTSC, and Wright-
Patterson AFB have resulted in requests to be kept informed of the work
since it applies to projects in their respective areas. The project has
been designed to permit application to microdonsitometry, color imagery,
electronic signals, etc., by anyone knowledgeable in the field. A follow-on
may prove desirable, but it would not necessarily be with the same contrac-
tor.
8. The sterility code is appro6priate; the Project officer will
assign security classifications to the individual reports.
25X1
25X1
9. It is requested that the negotiation with the 25X1
for a contract to conduct the program described ata cost not to excee
_~be approved..
ARTHUR C. LUNDAHL
Director
National Photographic Interpretation Center
Attachments:
1. Proposal
2. Form 21-120
APPROVED:
Assistant Deputy Director for Intell:i.gen0 D to
Distribution:
Original - NPIC/TSSG/SC&PS (After approval)
1 - ADDI
2 - NPIC/ODi.r
1,. - NPIC/TSSG
_l - NPIC/TSSG/RED
25X1
Approved For Release 2004/02/11 : CIA-RDP78BO5703A000200020021-9