5ER&
            UrritM)
            EnvirofVTMntBl Protection
            Agency
            Environmental Sciences Research
            Laboratory
            Research Triangle Park NC 27711
EPA-600 2-79-005
January 1979
            Bueerch and Development
Examination of
Automatic Data
Reduction  Methods for
Particle Field
Holograms

-------
                RESEARCH REPORTING SERIES

Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology.  Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:

      1.  Environmental Health Effects Research
     2.  Environmental Protection Technology
     3.  Ecological Research
     4.  Environmental Monitoring
     5.  Socioeconomic Environmental Studies
     6.  Scientific and Technical Assessment Reports (STAR)
     7.  Interagency Energy-Environment  Research and Development
     8.  "Special" Reports
     9.  Miscellaneous Reports

This report has been assigned to the ENVIRONMENTAL PROTECTION TECH-
NOLOGY series. This series describes research performed to develop and dem-
onstrate  instrumentation, equipment, and methodology to repair or prevent en-
vironmental degradation from point and non-point sources of pollution. This work
provides the new or improved technology required for the control and treatment
of pollution-sources to meet environmental quality standards.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.

-------
                                               EPA-600/2-79-005
                                               January 1979
    EXAMINATION OF AUTOMATIC DATA REDUCTION

     METHODS FOR PARTICLE FIELD HOLOGRAMS
                      by
                J.D. Trolinger
    Spectron Development Laboratories, Inc.
       3303 Harbor Boulevard, Suite G-3
         Costa Mesa, California 92626
            Contract No. 68-02-2491
                Project Officer

               Charles W. Lewis
  Atmospheric Chemistry and Physics Division
  Environmental Sciences Research Laboratory
      Research Triangle Park, N.C. 27711
 ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
      OFFICE OF RESEARCH AND DEVELOPMENT
     U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711

-------
                                 DISCLAIMER

     This report has been reviewed by the Environmental Sciences Research
Laboratory, U.S. Environmental Protection Agency, and approved for publication.
Approval does not signify that the contents necessarily reflect the views and
policies of the U.S. Environmental Protection Agency, nor does mention of
trade names or commercial products constitute endorsement or recommendation
for use.
                                      ii

-------
                                  ABSTRACT

     Holographic recording techniques provide one of the most powerful par-
ticle field diagnostic tools in existence.  A hologram can provide a frozen
three-dimensional image of a particle field through which detailed micro-
scopic examination of individual particles is possible.  Frequently, a par-
ticle field may contain many thousands of particles, and it becomes impracti-
cal for the human operator to glean all the data of interest from such a
hologram.  For holography to reach its full potential in particle diagnostics,
a three-dimensional image analyzer is required.

     The purpose of this study was to examine the feasibility of using exist-
ing electro-optic image analyzers to automatically analyze three-dimensional
image fields and to determine what modifications of existing equipment would
be required to construct such a system.

     Sample holograms as well as holograms produced in an actual field holo-
camera were used to make the evaluations experimentally, and well-refined
analytical descriptions of holographic images were used to add to the under-
standing of system requirements.  The study established that existing image
analyzers are capable within useful practical limits of locating particle
images in three-dimensions and measuring size and shape factors of the
particles.  A plan for integrating such equipment to produce a fully auto-
mated data reduction system is presented.

     This report was submitted in fulfillment of Contract No. 68-02-2491 by
Spectron Development Laboratories, Inc. under the sponsorship of the U.S.
Environmental Protection Agency.  This report covers a period from 9 March
1977 to 9 September 1977 and work was completed as of 1 January 1978.
                                     iii

-------
                                 CONTENTS

Abstract	ill
Figures	   vi
Tables	vil
Acknowledgement  	 viii

   1.  Introduction  	   1
            Background 	   1
            Purpose of the Report	   2

   2.  Conclusions and Recommendations	   3

   3.  Review of Particle Field Holography 	   4
            How Holography Offers Advantages 	   4
            Choosing a Holography System 	   6

   4.  Automatic Processing of Particle Field Holograms  	  11
            Introduction 	  11
            Existing Data Reduction Technique  	  12
            The Task of Automating Data Reduction	13
            Evaluation of the Quantimet System for Holographic
              Image Analysis	14
            Sample Data from Holograms	35

   5.  The SDL Field-Portable Particle Holocamera  	  41

   6.  A Proposed Automatic Holography Data Reduction System 	  45

References .	50

-------
                                  FIGURES

Number                                                                Page
  i-'-.v.-                                                                     ifO—

  1     Holography  of  a point  source  	  5
  2     In-line hologram of  a  particle field 	  8
  3     Off-axis hologram of a particle field  	  8
  4     Geometry for Equation  (1)	9
  5     Intensity distribution in the intensity of  an image	15
  6(a)  Hologram reconstruction setup  	  17
  6(b)  Experimental apparatus for machine data reduction of
          particle  field holograms 	  18
  6(c)  Adjusting the  Quantimet system for a data run	19
  6(d)  Recording hard copies  from the monitor	20
  7     Two  different  planes of focus from a three-dimensional
          reconstructed image  as displayed on the "Quantimet"   ....  21
  8     Definition  of  image  parameters 	  23
  9     Same two planes shown  in Figure 7  with all  information
          below a preset threshold rejected  	  24
 10     One  plane of the reconstructed three-dimensional  image
          of a dust particle field	26
 11     Two  centimeter change  in focus position of  the image
          field shown  in Figure 10	27
 12     Two  field different  focus planes in the droplet  	  29
 13     Fog  droplet field reconstruction 	  30
 14     Scanning in focus while processing with two level detection   .  31
 15     Two  focus positions  of a pair of 30 micrometer particles
          in a dust cloud  .	33
 16     Same particles as displayed in Figure 15 but shown here
          with two  level detection	34
 17(a)  Holographic size analysis - glass  beads  	  38
 17 (b)  Holographic size analysis - aerosol spray  	  39
 17 (c)  Holographic size analysis - dust storm	40
 18(a)  Field-portable particle holocamera 	  42
 18(b)  Side view of the holocamera installed behind a shield
          as applied in explosion diagnostics  	  43
 18(c)  Front view  of  the shielded optics  package	43
 19     Holocamera  reconstruction system 	  46
 20     Flow diagram of automated data reduction	47
                                     VI

-------
                                  TABLES

Number                                                                Page

   1  Printout of Information Taken from the Image Analysis
        of Figure 9(a)	22

   2  Printout of Information from Figure 9(b)	25

   3  Data for Figure 14(a)	28

   4  Data for Figure 14(b)	32

   5  Data Associated with Figure 16(a)  	 32

   6  Data Associated with Figure 16(b)  	 35
                                    ViX

-------
                               ACKNOWLEDGEMENT

     The author wishes to acknowledge the assistance of Mr. Brian Partridge
of the Cambridge Instruments Company in applying the Quantimet system to the
data reduction of  the holographic data.  Without his suggestions and assist-
ance, this study could not have been done.
                                      vi-ii

-------
                                  SECTION 1

                                INTRODUCTION

BACKGROUND

     Holography has provided a new and powerful technique in the study of
particle fields.  Holography can be used to freeze the three-dimensional
image of a particle field which covers a relatively large volume.  The holo-
gram is then used to reconstruct this three-dimensional image at which time
the image can be probed in great detail microscopically to study individual
particles and their properties, including size, shape, and velocity.  The
method can be used to extend the depth of field of the study of a dynamic
particle volume by three or four orders of magnitude over that available to
photography or microscopy.

     Although particle field holography has now been used for a number of
years in applications where no other method will work and has become an accept-
ed diagnostic tool, the method still suffers a number of serious limitations
which limit its use or make its use impractical for some studies.  Holograms
contain vast amounts of information.  Extracting that information can be an
extremely tedious and time-consuming task.  If that task can be performed
automatically by machines, then holography could be attractive in many more
types of studies than it is now.

     The field of electro-optics has produced a number of two-dimensional
image analyzing systems.  For example, an image can be inserted into such a
system by way of a photograph of a negative and the system almost instanta-
neously can convert this image information into a computer-compatible format
such that data processing is very quick.  Holographic images are three-
dimensional and, therefore, an image analyzing system must have the capability
of determining another important piece of information.  More specifically, the
system must be able to define which plane in the three-dimensional image is in
focus at any given time and is currently undergoing analysis.  All the infor-
mation taken from that given plane must be so designated to have arisen from
that particular plane.

     During this study we examined the capabilities of existing image analyz-
ing equipment to determine if such equipment was easily adaptable to provide
this requirement which was necessary for automated holographic data analysis.
We paid particular attention to a system produced by the Cambridge Instrument
Company Incorporated (IMANCO System) since the EPA currently has one of these
systems.  We found that this system already has the necessary programming
capability to determine when an image is in focus if the image is of suffi-
ciently high quality.  Therefore, our next step was to determine a quality

-------
factor which could be used to describe a hologram capable of producing such
an image.  The question,  therefore, became not can three-dimensional images
be automatically analyzed, but rather how good must the three-dimensional
image be in order for it  to be analyzed by currently existing image analyzers.
More specifically, how  good must  the hologram be which produced this image.
We then produced a number of holograms of both calibrated particle fields and
typical particle fields,  both in  the laboratory and in an SDL field-portable
holocamera, to determine  how practical such a complete system would be.  We
found that when the  proper conditions were met by correct field holocamera
design, it is entirely  feasible to produce useful holograms in the field which
can be analyzed automatically by  computer.  It is our belief that such a sys-
tem could be extremely  useful in  a vast number of requirements of the
Environmental Protection  Agency.

     A specific limitation of the system which may prevent  its widespread use
by a large number of groups at  this  time is the expense of  the overall system.
However, since the EPA  already  has the  image  analyzing computer,  this should
not be a factor for  the use by  the EPA.

     SDL has constructed  and used a  wide variety of  field-portable holocamera
systems under almost every conceivable naturally occurring  environmental
condition.  We have  found that  almost without  fail,  such  a  holocamera can be
tailored to meet the problems of  any naturally occurring  environment.  Such
systems have been used  in aircraft,  in  field  trailer systems,  and in weather
observatories,  in  conditions  ranging from  40°F below zero to  over 120°F,
under extreme weather conditions. These systems are capable  of producing
holograms  of quality sufficient to be reduced  automatically.   Such systems
will be  described  in this report.

PURPOSE  OF THE REPORT

     This  is a  final report  covering a  relatively small effort to determine
 the  feasibility  of  practical  use  of  automated  image  analysis  systems in re-
 ducing holographic  particle  field data.  In Section  3 we  provide  a brief
 explanation  of  particle field holography and  its variations after which the
 discussion is extended to the properties of reconstructed images.  This pro-
 vides  the  necessary background  for the  search  for an automated holographic
 data reduction  technique.  In  Section 4 we describe  an image  analyzing com-
 puter  system which was  chosen  for further  experimental work and describe
 experiments which were carried  out on actual  reconstructed  images from a
 variety  of holograms of particle  fields.   In  Section 5 we describe a typical
 field-portable  holocamera which can  produce holograms of  the  quality neces-
 sary for use  in an automatic  data reduction system.   In Section 6 a more
 detailed layout  of  the holocamera recording and data reduction system is
 presented.

-------
                                 SECTION 2

                       CONCLUSIONS AND RECOMMENDATIONS

     During the course of this study, we have produced a wide variety of
holograms.  These holograms range in quality from the ultimate or ideal qual-
ity that can be produced in a laboratory over a very limited known particle
field to the quality that arises from an operational portable field holocam-
era which is used in routine data collection in studies ranging from weather
to explosion diagnostics.

     Taking holograms produced for this study, we concluded that existing
optical image analysis equipment was sufficient to reduce images reconstruct-
ed from holograms produced in a good field holocamera.  Therefore, we have
concluded that the equipment necessary to automatically reduce holographic
data exists in various components and needs only to be assembled and inte-
grated for application to holography.  We have shown that the TV resolution
is sufficient for a wide range of interests of holographic images, the system
is capable of rejecting common types of noise which normally would make data
reduction difficult, the system is capable of eliminating common speckle
type of noise, the system is capable of sizing particles of interest, and the
system is capable of determining when a particle is in focus.

     The remaining task is to design such an integrated system in detail so
that the remaining components can be procured and assembled into a working
unit.  At this stage the computer software for that specific system must be
developed.  The principal item to be procured in this development is the
computer-driven x,y,z traverse onto which the hologram is mounted.  This
traverse could be driven by a number of different computers.  In fact, the
Hewlett-Packard desk top calculator which we used during this study could
perform that task, although it would be limited in speed.  A more preferable
system would be a minicomputer which would allow the data handling as well
as the traversing capability to be greatly enhanced.  We would recommend the
EPA make plans to assemble such a system for use with their existing
Quantimet 720 image analyzer such that an automated holographic data analyz-
ing system would be available for use in particle field studies.

-------
                                  SECTION 3

                     REVIEW OF PARTICLE FIELD HOLOGRAPHY

HOW HOLOGRAPHY OFFERS ADVANTAGES

     The purpose of this  section  is  to review the  current state-of-the-art
and to present examples of applications of holography  in the study of parti-
cle fields.

     An optical hologram  is a special type of recording of  a light wave which
stores all the information necessary to reconstruct  an essentially identical
light wave at a later time.  The  three-dimensional image of a  field  is viewed
through a hologram as if  the hologram were a window  behind  which  the actual
field was frozen in time.  It is  a simple matter to  convert this  virtual
image (which lies  behind  the hologram) to a  real spatial image (which lies on
the viewer's side  of the  hologram).  A hologram is formed by mixing  coherent
light scattered from the  field of interest  (object beam) with  a reproducible
mutually coherent  (reference) beam,  Figure l(a), and recording the intensity
of the sum on a photographic emulsion or other  recording medium (there are
many types) .  When the hologram is once again illuminated by the  reference
beam, Figure l(b), the hologram,  acting as a special type of diffraction
grating, changes the reference beam  into several components, one  of  which is
nearly identical in information content to the  original object beam.  The re-
constructed beam is never exactly identical  to  the original beam  simply be-
cause the recording and reconstruction processes are not ideal.   Optical
noise and aberrations cause resolution limits on the process which are thor-
oughly discussed,  for example, in Reference  1.

     When particles can be trapped on slides, replicated in plastic, or
frozen in space, conventional methods of viewing and photographing are usu-
ally satisfactory, if not superior.  When they  cannot  be trapped, more uncon-
ventional methods  are necessary.  The need for  in  situ characterization of
particles in all fields of science has spurred  the development of many tech-
niques based on light scattering  and extinction laws,  nucleation  phenomena,
shadow detection,  and measurement of other properties  which can be related
back to particle size, number, and type.  In many  cases large  numbers of
particles are involved and statistical properties  of the entire sample are of
more interest than individual particles.  In such  cases the detailed examina-
tion of individual particles may  not be required.  Holography  is  particularly
powerful when it is desired to examine individual  particles in a  distribution.

     All optical imaging  measurements, holography  included, are resolution
limited by the laws of diffraction.  The precise limit of resolution for any
optical sy^rem depends upon a somewhat arbitrary definition of resolution;

-------
   Liser Light
    (a)   In-line and off-axis hologram
         of a point  object.
                 Off-Axil Hologram ~B"
                                      Rnllitilgi
    (b)   Forming the  image with  the
         hologram.
Figure  1.   Holography of a point  source.

-------
however, the same
-------
     An in-line hologram is made by passing coherent light through a volume
of particles and recording the resulting wavefront (Figure 2).  Part of the
light is diffracted by the particles and the remainder passes through the
field unscattered.  The unscattered (reference) light interferes with the
scattered (object) light and is recorded on a photographic medium.  An off-
axis hologram is formed by introducing a reference wave which is separate
from the light passing through the object field onto the recording (Figure 3).
The laser beam is split initially in two.  One portion illuminates the object
field, while the other serves as the reference beam.  Many geometries are
suitable for off-axis holography.  The object field can be illuminated dir-
ectly with the laser beam or with laser light that has been diffused by a
frosted glass or similar diffusing medium.

     The object wave in both in-line and off-axis cases is reconstructed by
illuminating the developed recording with the reference wave.  Equation (1)
predicts the position of the reconstructed image as a function of the wave-
length, radius of curvature, and position of the reference wave.  (Figure 4
illustrates the geometry for Equation (1).)

-1
I1 + "2 ¥
\z Xn z
\ p 1 r
M'1
A-z I
1 o/
                              A  z     +X2z       z
                     x.   =  +-	x  -T	x+ — x                  (1)
                      a.        Xn z   o   X.. z   r   z   p
                               1  o       1  r       p  v


                              X? z.      X_ z.      z.

                     yi   =  + \:~ry0 ~ x7^yr ^v
                               1  o       1  r       P

These equations result from a first order analysis of thin holograms and are
not accurate when large  changes in wavelength or reference wave curvature
are encountered.  A more general analysis is treated in Reference 1.

     There are no hard,  fast rules for choosing one specific type of system
for a given application.  In-line holography is the simplest type of holog-
raphy and perhaps should always be considered first in particle field studies.
However, there are many  cases in which off-axis holography is applicable and
in-line holography is not.  Such cases exist when the particle number density
or their total cross section is very large or when the object field signifi-
cantly modulates the phase of the electromagnetic radiation.  A sufficient
amount of light (typically about 80 percent) must pass through the field with-
out modulation to serve  as a reference wave to produce a good in-line
hologram.

     When a field is illuminated by diffuse illumination, the resulting re-
constructed image is easier to view with the naked eye.  Associated with
diffuse illumination are speckle patterns which cause problems in high reso-
lution viewing.  In any  viewing plane the background of illumination is not

-------
    Direct _	

    Illumimtion
                      Scittend Light
                                             - Hologram
                                             - Interfmnx
                                              bemem
                                              ScittnedMd
                                              Unscattereif
                                              Light
Figure  2.   In-line hologram of  a particle field.
                                                Holognm
Figure  3.   Off-axis hologram of  a particle field.

-------
                                             +y
      x y z
       ooo
                                                               Hologram
                              Producing the Hologram of a Point
sub o ~ object wave
sub r ~ reference wave
sub p — reconstructing wave
sub 1 ~ reconstructed wave
     x y z
      P P P
                               Reconstructing the Point Image
                 Figure 4.   Geometry for Equation (1).

-------
uniform; it is formed by speckle points  of varying  size and  separation.  The
mean speckle separation is about twice the resolution of  the original  record-
ing system.  Therefore, the resolution limit  of  the holography  system  must  be
five to ten times smaller than  the  smallest particle to be observed  so that
particles can be distinguished  from speckle.

     To see the smallest particles  in the reconstructed image requires a
total system of low noise.  Even those particles which are much smaller than
the system resolution limits  can be seen as poii'its  of light  whose brightness
can be associated with size.  The task is to  separate these  points from the
optical noise.  One of the most powerful methods for seeing  particles  in the
submicron and micron size range is  the holographic  recording of scattered
light only, against a dark background.   The particle field is illuminated by
a pencil' or sheet of light and  the  hologram of scattered  light  is made at
some angle away from the axis of the illuminating beam.

     The hologram itself is often the poorest quality optical element  in the
system because the emulsion adds phase and amplitude noise to the recording
and reconstruction process.   This element can be improved by such techniques
as liquid  gating and special  photographic procedures^.  Such methods are
tedious  and are avoided in most applications.

     Holograms possess most of  the  properties of lenses.  In practice,
however, lenses are better imaging  elements than holograms and  are combined
with  them  to  attain higher resolution and larger working  distance.   When the
 field of interest must be located at a relatively large distance from  the
nearest  optical element, lenses or  mirrors are normally used to image  the
 field  closer  to the hologram  to maintain high resolution.  A lens hologram
 combination can be used in such a way to completely correct  aberrations in
 the  lens during reconstruction. Although high magnification is possible in
principle  in  holography without lenses,  the most successful  applications have
 resulted from combinations of conventional with  holographic  magnification.
A well-designed holography system will normally  produce ultimate image qual-
 ity  and  resolution which is slightly poorer than a  conventional optical sys-
 tem which  is  focused on an object plane, but  it  will do this over a  volume
 in  a  single recording.  A detailed  review of  particle field  holography is
 given  in References 2  and 3.
                                      10

-------
                                  SECTION 4

              AUTOMATIC PROCESSING OF PARTICLE FIELD HOLOGRAMS

INTRODUCTION

     There are two basic limitations of holography systems for particle field
studies which are suffered in almost all current applications.  First, data
recording rates are too slow.  One sample volume, typically on the order of
a liter, can be recorded every few seconds.  This rate is too slow to provide
ample statistics in many cases.  Second, data reductions from holograms re-
quire tedious, time consuming procedures which make it difficult to reduce
even the quantities of data generated by existing systems.

     The technology for increasing data recording rate is here; however, it
would make little sense to increase the data recording rate without first in-
suring that methods to rapidly reduce the data are forthcoming.  Therefore,
it is reasonable to first remove the holographic data reduction bottleneck
and define the quality of data required for rapid data reduction.  This will
tell us how good the recordings must be and how many can practically be
reduced.

     Two different kinds of holography data must be considered, depending upon
the type of light which is used to illuminate the particle field.  The field
may be illuminated with diffuse light or with direct laser light.

     To illuminate a particle field diffusely, the laser beam is expanded onto
a frosted glass or other diffuser which is placed behind the sample volume.
A reference beam is carried around the sample volume where it is mixed with
the light from the sample volume onto the hologram.  The hologram looks like
a uniformly exposed piece of film, appearing to contain no information at all.
The reconstructed image appears to contain dark, shadow images against a
diffuse light background.  As previously explained, the problem with this
method occurs in high resolution studies.

     Diffuse illumination has some advantages.  The reconstructed image can
easily be viewed with the naked eye, with the particle field appearing against
a bright background.  A particle is illuminated from a relatively wide angle
of directions, and accordingly can be viewed over that wide angle.  This
provides two benefits.  First, since forward scattered light is greatest, it
provides for a wider angle over which forward scattered light can be recorded.
This makes a more uniform recording which would exhibit higher resolution,
except for the effects of speckle.  Second, the particle image, being backed
by a diffuse field, has essentially no out-of-focus shadow.  The out-of-focus
image blends into the background in such a way that it cannot be distinguished


                                      11

-------
from the background when defocused by typically 100 particle diameters
(depending upon the system).  This feature provides a simply defined  focusing
property of the image.

     Direct light holograms are produced when the particle field is illumi-
nated by light directly from the laser with lenses being used  to expand  the
beam large enough to fill the sample volume.  The reference beam is either
(1) the unscattered beam passing through the sample volume or  (2)  a separate
beam introduced onto the hologram by another path around the sample volume.
In either case, the hologram appears as a more or less uniformly exposed
piece of film with diffraction patterns and fringes superimposed.  For each
particle in the volume from the laser to the film there is a diffraction
pattern (a coherent shadow) which is a set of more or less circular concentric
rings.  When reilluminated by the laser, each of these diffraction patterns
diffracts light into a particle image in its original position in  3-D space.

     When passing through the focused plane of an image, the diffraction
pattern collapses into a sharp high contrast image which is bright against
dark for in-line holography and dark against bright for off-axis holography.
After passing through the focused plane, the defocused image (shadow  or  dif-
fraction pattern) expands in size and loses its characteristic shape  ulti-
mately appearing as a somewhat circular set of diffraction rings bearing no
resemblance to the original particle.

EXISTING DATA REDUCTION TECHNIQUE

     The reconstructed 3-D image of the sample volume is scanned either  by
moving the hologram and passing the image over a viewing device or by moving
the viewing device through the sample volume image.  The simplest  type of
viewing device is a microscope which the viewer scans by hand.  This  technique
is the one we use in the field for hologram quality evaluation.  However,
viewing large numbers of reconstructed holographic images through  a microscope
by eye is extremely tiring.  In most laboratories, extensive data  reduction
is accomplished by converting the coherent reconstructed image (normally a
red image from a HeNe laser) to an incoherent black and white  image by pro-
jecting it into a closed circuit TV system.  Final viewing of  the  image  is
done on the TV monitor.  When focused on the TV monitor, the image may be
studied, sized, photographed, or sketched.

     Consider a typical particle field hologram from our field portable  holog-
raphy system.  The sample volume has a cross sectional area of about  25  cm2
and a length of 15 cm.  A 100 micron particle should appear about  1 cm across
on the TV monitor for comfortable viewing; that is, a magnification of about
100.  The TV magnifies electronically by 25, the remainder is  achieved opti-
cally upon reconstruction either with lenses or holographically (see  Equation
(1)).  The actual cross sectional area viewable on the TV screen represents
an actual area of about .05 cm2 under these conditions.  Therefore, 500  in-
depth scans are required to analyze the entire hologram.  This can take  an
operator an entire day, depending upon what data he is collecting.

     Considering the fact that in a single day's operation of  the  system hun-
dreds of holograms can be recorded, the data handling problem  becomes

                                     12

-------
cumbersome.  Normally the data output comprises photographs, number  counts,
size distributions and, in some cases, spatial and velocity distributions.

THE TASK OF AUTOMATING DATA REDUCTION

     Consider the task performed by the human operator.   In general  he must
scan through the 3-D field while watching image characteristics.   In the  case
of direct light holograms, he can tell directly from  the  hologram  what the xy
positions of particles will be by locating diffraction patterns.   Some of
these particles are contamination on windows and  in the optical  train.  If
two holograms made in the same system at different times  are placed  side  by
side, he can mentally subtract one from the other leaving only diffraction
patterns which are not common to the two.  These  represent the particles  in
the sample volume.  He could immediately count the number of particles (or
determine velocities) directly from the hologram  without  reconstruction.  To
see the particle he must locate the focused position, z,  during  reconstruction.
The human operator uses the following characteristics to  decide  upon the  fo-
cused position:

     (1)  The diffraction pattern converges to form an image;
     (2)  The light distribution becomes more intense in  the image;
     (3)  Image edges become defined;
     (4)  He sees an expected shape or combination of shapes emerge;
     (5)  Optical noise is mentally rejected because  it has different
          characteristics than the image.

     Every criteria used by a human operator can  be stated mathematically and
can be sensed optically and/or electronically.  Therefore, in principle,  the
reduction of holographic data can be done by machine.  In fact,  for  almost
every criteria there exist several machine processes which can,  in principle,
perform the human task.  Determining factors in the rapidity with  which a
human operator can function in holographic data reduction are memory and
response to learned characteristics.  Both factors are poor during the first
hour or so of a data reduction session.  They peak for a  period  of only an
hour or so and then decline as the operator becomes tired or succumbs to
monotony,  resulting in variable judgment and poor efficiency.  A well-pro-
grammed machine would excel in this respect because it could exercise the best
knowledge  of a skilled operator with more speed,  consistency and endurance.

     A higher quality hologram requires less skill and judgment  in data re-
duction and, therefore, can be automatically reduced by a less complicated
machine.  As the quality of a hologram get poorer, so does the practicality
of devising an automatic data reduction machine.  (In fact, a certain point
is reached whereupon the required level of expertise  of a human  operator
increases.)  One of the tasks in automatic data reduction is to  determine
practical machine limitations for performing the  above tasks so  that a mini-
mum quality hologram can be specified.

     During this program, we examined a number of possible methods to accel-
erate data reduction.  The most promising immediate approach was the exten-
sion of existing two-dimensional image analysis techniques.  We  performed
experiments with the IMANCO "Quantimet" image analyzer located at  the

                                     13

-------
Cambridge IMANCO Instrument Company, Mountain View, California,  office.   The
results are described in the following subsection.

EVALUATION OF THE QUANTIMET SYSTEM FOR HOLOGRAPHIC IMAGE ANALYSIS

Existing Work

     Shortly after we had begun this investigation for  the  Environmental
Protection Agency, a group of scientists with whom we had been in  touch  work-
ing for the Meteorological Office in the United Kingdom had independently
chosen the IMANCO system to apply to the data reduction of  holograms  produced
in weather and in environmental studies.  In Reference  4, they describe  a
series of experiments in which they analyzed holograms-with this*system.   The
information gained from those studies actually gave us  a head  start in the
conduct of our own studies.

     As previously described, the new problem involved  in adapting the IMANCO
system to holographic data reduction is to  force  this system to differentiate
between particles which are in focus and out of focus,  therefore,  allowing a
scan automatically of the third dimension'.  They  showed, for example,  that the
existing IMANCO system was capable of performing  the first  three tasks which
are outlined in the previous subsection.  Consider, for example, the  intensity
distribution illustrated in Figure 5, which represents  the  cross-sectional
intensity in the  image of a particle as the particle passes through focus.
Three levels were set into the IMANCO system to be used in  image detection.
Level 1 was set to be greater than the normal background noise.  The  IMANCO
system, therefore, rejects any optical information below this  level and  does
not display it on the TV monitor.  Level 2  is set at some value higher than
Level 1 but lower than the maximum intensity of an in-focus image  of  the
particle.  Level  3 was set to be higher than the  intensity  in  the  in-focus
image of the particle.  The determination of focus is made  in  the  following
way.  When the image intensity exceeds Level 1 but not  Level 2,  it can be
assumed that the  particle is very badly out of focus and the system can  con-
tinue scanning toward focus.  When the intensity  in the image  increases  to
the point where it exceeds Level 2, it can  be assumed that  the image  is  ap-
proaching its in-focus condition and the scanning system can be slowed down.
When the image nears its in-focus condition, the  diameter of a particle  as
determined by the extent of the image going out to Level 2  should  be  approx-
imately the same  as that determined from Level 1.  When the difference be-
tween the Level 1 diameter and the Level 2  diameter is  at a minimum,  the
particle can be considered in focus, and the size determined by an average of
the Level 1 and Level 2 diameter should be  a good representation of the par-
ticle size.  That information is recorded for that particular  particle image.
For some cases, an out-of-focus condition produces a higher intensity at  the
center of the particle image than away from the center, and these  cases  are
rejected when the intensity in the image exceeds  Level  3.   The IMANCO system
has the correct programming to allow the determination  of minimum  values  such
as that described here, and an output is available to drive interface equip-
ment which would, in turn, control traverse systems.  During the reported
studies, a Hewlett-Packard 9830A was interfaced with the Quantimet to  perform
the task of driving the traverse which held the hologram under study.
                                      14

-------
                                           Intensity
                                                                  Level. 3
  Portion of Inag*
Betw»«n Lavels 162
                                                         Radial Distance
                     (a)  In-Fooxm Im»«« intensity Distribution.

                                           Intensity
                                                                  Level  3
                                                                  Level  2
                                                                  Level 1
                                                          Radial Distance
                                 (b) Out of Focus.
                                           Intensity
                                                                  Level 2
                              (c) Badly Out of Focus.

     Figure  5.   Intensity  distribution  in the intensity  of an  image.

-------
Experimental Investigation

     To further evaluate the ability of  the  "Quantimet" system in hologram
image analysis, we produced a wide variety of holograms.  Three basic  types
of holograms were produced: (1) ideal laboratory quality under clean condi-
tions (these were intended to represent  the  best case recordings),  (2)  holo-
grams produced in the SDL field holocamera of controlled particle fields
(controlled size and number density), and (3) holograms produced  in the SDL
field holocamera in a variety of natural environments.

     Our intent was to determine what limitations would be  exhibited by the
"Quantimet" system.  At first, we had planned to begin with the ideal  best
case holograms to provide a baseline.  We decided very quickly that this
would not be necessary.  We observed that the Quantimet could, in fact, ana-
lyze almost all of the holograms we had  produced.   Some limitations will  be
illustrated.  Therefore, we advanced to  the  holograms produced in the  field
tests, spending most of our time using the Quantimet to analyze actual field
recorded holograms of particles.

     The system used was the Cambridge IMANCO Quantimet 720 image analyzing
computer interfaced with a Hewlett-Packard 9830A desk top calculating  computer.
All experiments with the Quantimet were  conducted at the Cambridge  Instrument
Company's Mountain View office with  the  assistance  of Mr. Brian Partridge, a
Cambridge employee.  The required holographic reconstruction equipment was
transported to Mountain View for the experiments.

     Figure 6(a)-(d) shows the experimental  apparatus.  The reconstructing
laser was a Spectra-Physics Model 124 HeNe laser producing  15 mw of laser
power at 633 nanometers wavelength.  This beam was  expanded and collimated
after which it illuminated the hologram.  All holograms were produced  on  4x5
inch Agfa-Gavaert, Scientia emulsions.   Both Agfa 10E75 and 8E75 plates were
examined.  Both in-line and off-axis recordings were produced.  Only the  in-
line recordings were examined since  a larger laser  is required for  off-axis
reconstruction.  An adjustable neutral density filter allowed us  to adjust
laser power at the hologram to optimize  the  interface between the reconstruc-
tion system and the TV camera.

     The reconstructed image produced by these holograms is a one to one  re-
construction of a 50 mm by 150 mm cylindrical volume which  starts 10 mm from
the hologram and extends to a distance 160 mm from  the hologram.  This  image
was magnified and relayed into the TV camera by a high quality copy lens,
specially selected for this problem.  As pointed out in Reference 4, cleanli-
ness is an important factor with all optical elements in this system.

     Figure 7 shows the face of the  Quantimet monitor when  a reconstructed
three-dimensional image is presented to  the  associated TV camera.   This is
the "raw" image with none of the image processing capability engaged.   Two
different focus planes are illustrated.  These figures can  be used  to  explain
the meaning of the various features  in the figure.  The inner frame is  the
active region for the system.  The system can be described  as a matrix of
picture points 800 points wide by 625 points high.  Therefore, it is conven-
ient to use dimensions of picture points.  In the top figure, a dimensioning

                                      16

-------
                                                                          TV Camera
                   Adjustable
                      .  mt*r
Hewlett-Packard
   Desk Top
  Calculator
 IMANCO
Analyzer
             Figure 6(a).   Hologram reconstruction setup.

-------
-
                       Figure 6(b).  Experimental apparatus for machine data reduction
                                     of particle field holograms.

-------
-
                         Figure  6(c).   Adjusting the Quantimet system for a data run.

-------
I
<
                                           Recording hard copies from the monitor.

-------
                       x - From 0 to 800 Picture Points
    y - From
    0 to 625
    Picture
    Points
                (a) z = 5 cm
NOTE: The hologram was made in an SDL field-portable holocamera in heavy  fog.
     Figure 7.   Two different planes of focus from a three-dimensional
                reconstructed image as displayed on the "Quantimet."

-------
raster has been displayed on the monitor.   The  small  elements  are  spaced by
10 picture points  (p.p.).  The number  in  the  top  left corner displays  how many
features (particles) were counted  (six for  this case).   The top  righthand num-
ber is the last dimension measured by  the system.
     We have identified certain regions  of  the  image  to  illustrate its appear-
ance as particles pass through focus.  Region A  contains  a diffraction pat-
                                                3-
tern which is the out-of-focus coherent  image of  the  particle shown in region
A, .  Region B  contains an in-focus  image which is  out of  focus in region B, .

These images are 20 micron diameter  water droplets.

     The program which we used determined the x and y positions of each fea-
ture, its area (A), perimeter (PERI), horizontal  feret  (FH), vertical feret
(FV), 45° feret (FE), 135° feret  (FW), horizontal projection (PH), vertical
projection (PV), ratio of area to perimeter, and  ratio of  area to  perimeter
squared.  Figure 8 illustrates the definition of  each of these dimensions.

     When the image  is first presented to the TV  camera, it appears as in
Figure  7.  The noisy background characteristic  of holographic images is
clearly evident.  The first processing step is  to set a  threshold  value of
intensity below which all information is rejected.  Figure 9 shows the images
resulting from Figure 7 when this is done.   This  rejects entirely  the small
out-of-focus  images whose intensity  is spread over  the Fraunhofer  diffraction
pattern.  See, for example, regions  B and  A^.  Larger particles are still
                                      3.       D
detected over a rather large depth of field as  can  be expected.  However, so
long  as size  and not position is  needed, the size of  the particle  image does
not change drastically for a quite large depth  of field, so it could be meas-
ured  with reasonable accuracy in  either  figure  (although it is truly in focus
in  only one  thin planar region).

      In Figure 9(a)  six features  were counted,  while  in  Figure 9(b)  only five
were  counted.  A marker appears on the monitor  beside each feature which is
being analyzed.  This will be useful to  an  operator allowing him to quickly
evaluate the  validity of what is  being analyzed.  A typical printout of data
is  shown in  Table 1.   (Dimensions are given in  picture points, p.p.)
         X
TABLE 1.  PRINTOUT OF INFORMATION TAKEN FROM
          THE IMAGE ANALYSIS OF FIGURE 9(A)

   AREA  PERI  FH   FV   FE   FW   PH   PV
A/P
A/P'
1.
2.
3.
4.
5.
6.
235
241
473
678
482
321
163
273
355
526
643
670
101
154
194
79
34
485
39
47
56
34
23
82
11
15
18
10
8
26
12
14
17
10
8
24
16
18
18
14
8
30
14
18
20
14
12
30
12
14
17
10
8
24
12
15
18
10
8
26
2.590
3.277
3.464
2.324
1.478
5.915
0.066
0.070
0.062
0.068
0.064
0.072
 The  particles  actually  ranged  in  size  from about  10  to  50  microns.
                                      22

-------
                                              Vertical
                                               Feret
          Horizontal Feret
       Vertical Projection = a + b
                                             Horizontal
                                             Projection
Figure  8.   Definition of image parameters.
                     23

-------
No. of Features
   Processed
     Last Programmed Process  (1)
Performed and the Measured Value  (93)
                   z = 5 cm
                    z =  7  cm
      Figure  9.   Same  two planes  shown  in Figure  7 with all information
                 below a preset threshold rejected.


-------
     When we changed the focus position to  that leading to Figure 9(b),  the
following printout resulted.

             TABLE 2.  PRINTOUT  OF  INFORMATION FROM FIGURE 9(B)

        X     Y    AREA  PERI  FH    FV    FE   FW   PH   PV    A/P     A/P2
1.
2.
3.
4.
5.
283
286
521
456
366
162
275
353
516
674
315
310
227
11
596
53
64
55
13
92
13
20
18
4
28
14
20
16
3
26
16
24
22
8
34
16
24
22
4
30
16
20
16
3
26
16
20
18
4
30
5.943
4.844
4.127
0.846
6.478
0.112
0.076
0.075
0.065
0.070
     Comparing the data  in Tables  1  and  2,  the  first  two  particles  are  the
same.  The particle diameters measured vary by  15  percent and  25 percent  in
horizontal feret respectively.   The  nearness in equality  of vertical  and
horizontal feret are  characteristic  of spheres.  The  third particle is  the
same in both figures  and so  are  the  measured parameters to within a few
percent.  The fourth  particle is different  in the  two figures  and size  is the
reason.  The fourth particle image in Figure 9(a)  is  small and did  not  carry
enough intensity into the camera to  be above the threshold in  the setting for
Figure 9(b) where it  was considerably out of focus.   The  same  is true for
particle No. 4 in Figure 9(b) and  No. 5  in  Figure  9(a).   The sixth  particle
in Figure 9(a) is the same as No.  5  in Figure 9(b).   This is the largest
particle in the set.   Its size is  determined within 7 percent  the same  for
the two cases.  In tb.e actual setup  the  true focus position of these  images
could be easily seen  to  within better than  one  millimeter, though it  is not
completely evident in the photographs.

     The key to successful implementation of such  a system in  holographic data
reduction is its capability  to handle noise.  Figures 10  and 11 illustrate
some of the studies in noise characteristics and system response.   Several
kinds of problems are illustrated  here;  and one by one we discuss their re-
moval by the system processing functions.   In Figure  10 the intensity across
the image varies.  When  this variation is severe,  picture "shading" correc-
tion capability can compensate for this  effect.  We did not find this neces-
sary on data we were  analyzing.

     Speckle noise is normally a serious factor in image  processing.  We
found in every case examined that  speckle noise could be  almost entirely re-
jected by the proper  setting of  a  so-called resolution limit.   Speckle  noise
is made up of intensity  distributions of extremely small  "points" of  light.
When the speckle size is much smaller than  particles  of interest, it  can be
rejected by existing  instrument  functions  (compare Figure 10(a) and 10(b».

     The two images in the bottom  right  were circled  by diffraction rings,
even after processing.   Again this effect is caused by the brighter image on
the righthand side of the screen and can be removed by shading correction.

     We studied the focusing problem and the possible system functions  which
can be used to drive  an  automatic  scanning  system  to  focus.  We found several
                                      25

-------
       (a)  Entire unprocessed image as presented  to
            the image analyzer.
       (b)  Image after rejecting all information below
            a preset threshold.


Figure 10.  One plane of the reconstructed three-dimensional
            image of a dust particle field.
                             26

-------
       (a)  Unprocessed.
      (b)  Information above a preset  threshold.
Figure 11.  Two centimeter change  in  focus  position of  the
            iraage field shown in Figure  10.


-------
ways to achieve this, but one in particular was  rather simple and existed in
the normal system programs.

     Figure 12 shows recordings at two different reconstructed  image  planes.
The system was set to examine that part of the image which  lies between two
intensity levels.  The result is shown in Figure 13.  It  can be seen  that
each image is now represented by an annulus of light.  This annulus of light
associated with the original images varies much  more rapidly with z position
and the properties varying are easier to quantize  than the  entire image.
     Figure 14 includes some interesting comparisons at two z positions for
several particles.  The image marked A  is present as an annulus in Figure
                                      cl
14(a).  In Figure 14(b) this image (marked A.) is out of focus 2 cm.  The

intensity in the image reaches Level 1 but not Level 2.  Therefore, it is
shown as a solid image.  The image marked B^
14(a) than it is in Figure 14(b).
increases in Figure 14 (b).
                                             is  closer  to  focus  in Figure

                                   The  thickness of  the annulus,  therefore,
      Tables  3  and  4  are  the printout  data  for  these  two  cases.

                       TABLE 3.  DATA  FOR FIGURE  14(A)

        X     Y    AREA  PERI   FH   FV   FE    FW  PH    PV    A/P
    1.   379
    2.   253
    3.   254
    4.   492
    5.   489
    6.   118
    7.   675
    8.   691
    9.   338
        324
        343
  10.
  11.
  12.  351
  13.  356
  14.
  15.
        305
        316
   16.
   17.
   18.
   19.
       356
       308
       354
       331
80
160
274
347
353
375
460
518
637
639
640
645
656
662
663
663
671
671
674
17
26
70
17
11
23
12
7
18
30
8
1
22
8
28
17
20
26
120
22
52
86
31
23
29
26
4
21
24
12
4
23
12
47
16
22
21
172
5
10
15
9
10
6
6
1
7
9
5
1
5
4
12
4
5
6
27
4
9
16
10
16
6
5
2
3
6
2
1
8
2
17
15
7
7
29
8
14
18
6
12
8
8
2
8
12
4
2
10
4
20
10
6
10
32
8
14
18
12
16
8
6
2
6
8
4
2
10
4
8
16
10
8
26
6
17
27
10
5
8
7
2
4
5
2
1
8
2
17
6
7
7
36
7
17
28
10
9
9
10
1
8
9
5
1
6
4
13
4
7
6
42
0.773
0.500
0.814
0.548
0.478
0.793
0.462
1.750
0.857
1.250
0.667
0.250
0.957
0.667
0.596
1.063
0.909
1.238
0.698
0.035
0.010
0.009
0.018
0.021
0.027
0.018
0.438
0.041
0.052
0.056
0.063
0.042
0.056
0.013
0.066
0.041
0.059
0.004
      The  third particle  in Table  3  is  the  first  in Table  4.  We will  describe
 each  measured parameter  in more detail in  a more appropriate figure later.
 However,  these figures show other important points.   In Table  3,  19 features
 (particles) are described.  Starting with  particle No. 9  through  19,  all
 particles are very close to the same x-y position  (i.e.,  around x = 320,
y = 650).  These are all, in fact, describing
                                                   region marked  C  .   The
                                      28

-------
Figure 12.  Two field different focus planes in  ths droplet.


-------
(a)   Unprocessed image presented to the image
     analyzer.
(b)   Only intensity lying between two gray levels
     is displayed and processed.
  Figure 13.   Fog droplet field reconstruction.
                       30

-------
             (a)  z = 0
             (b)  z = 2 cm
Figure 14.  Scanning in fomc while  processing  with twu level aetection.


-------
large diffraction ring had broken into what was  interpreted as,particles.
This "erroneous" result was deliberately  left  in this  picture for purposes of
illustration.

                      TABLE 4.  DATA FOR  FIGURE  14(B)
                                                                          2
        X     Y   AREA   PERI  FH   FV    FE    FW   PH    PV    A/P     A/P
1. 356
2. 594
3. 839
4. 722
5. 440
274
356
401
618
646
137
122
18
22
4
73
113
28
35
10
17
19
6
6
4
18
21
5
8
1
22
24
10
10
4
22
24
8
10
4
22
37
7
13
1
25
37
10
11
4
1.877
1.080
0.643
0.629
0.400
0.026
0.010
0.023
0.018
0.040
     The diffraction ring can be rejected by the  several methods  mentioned,
or it can be removed by the operator using a light pen.  Any area on the
screen which appears objectionable by an operator can be circled  and removed
entirely from the processed data.  In Table 4  this particular region was
moved outside of the detected area and rejected.

     Figures 15 and 16 illustrate in more detail  the properties of an image
when passing through focus.  We picked two reconstructed images of dust
particles and magnified them to show more detail.  They lie  in close vicinity
in x and y, but are considerably separated in  z  (about 1 cm).  Figure 15
shows the unprocessed images.  The particle size  was about 30 microns.   The
top particle is in focus in Figure 15(a) while the bottom is in focus in
Figure 15(b).

     Figure 16 illustrates the system output when the two intensity levels
are set and detection limited to between the levels.  Comparing Figures  15
and 16 clearly shows how much better the focus position can  be found when
the annulus is displayed instead of the entire image.  This  is related to the
fact that the gradient of the intensity at the image edge is what is used by
an observer in finding focus.  The thickness of the annulus  is a  direct  meas-
ure of the intensity gradient at the image edge and, therefore, provides the
operator more direct information.  In the usual case, an operator must look
at the intensity at the image edge and mentally determine changes in gradient.
This can be done only if the image is being scanned, a reason why it is  hard
to illustrate in a report.

     Tables 5 and 6 describe the processed images in Figure  16.

                 TABLE 5.  DATA ASSOCIATED WITH FIGURE 16(A)

        X     Y   AREA   PERI  FH   FV   FE    FW   PH   PV     A/P     A/P2

   1.  386   330   487    455  57   70   72    66   150  152   1.070   0.002
   2.  549   621  1142    397  56   60   60    64   109  133   2.877   0.007
                                      32

-------
        (a)  Top particle  in  focus.
      (b)  Bottom particle in  focus.
Figure 15.
Two focus positions of a pair of  30 micrometer
particlco in a ciuat cloud.


-------
         (a)   Top particle in focus.
        (b)  Bottom particle in focus.
Figure 16.  Same particles as displayed in Figure 15 but shown
            here with two level detection.
                                -

-------
                 TABLE  6.   DATA ASSOCIATED WITH FIGURE 16(B)

        X     Y   AREA    PERI  FH   FV   FE_   FW   PH_  PV_   A/P     A/P2

   1.  494    349  1476     644  72   101  78   82   121  220  2.292   0.003
   2.  623    629    421     420  58    56  70   58   119  144  1.002   0.002

Here we describe in detail the signficance of the measured parameters.

x,y Positions—
     The intensity  distribution (particle image) numbered as 1 and 2  corre-
sponds to the same  particle in both figures.   The x and y positions would  be
identical if  the z-scan was precisely normal  to the TV vidicon.   Since  it
was not, the  positions  on the screen vary somewhat (more in x than in y).

Area—
     The area of the annulus was thought to be a direct measure  of in-focus
position.  We found this partially so, but we did observe cases  where area
actually gets smaller outside of focus.  This is a property of the diffracted
light field of the  particle.

Perimeter—
     Perimeter, like area, can get smaller outside of focus, especially for
a non-spherical particle with complex edge detail.  Particle No.  2 is an
example.

Horizontal, Vertical, 45°, 135° Feret—
     These are the  same for a spherical particle.  As shown here,  they  de-
scribe quite  well the particle shape.

Ratio of Area to Perimeter—
     We found this  ratio to be an excellent guide to determine focus  position.
Recalling that the  dimensions are in picture  points, this ratio  would equal
unity when the annulus  is two picture points  wide because the area is equal
to  the number of picture points in the annulus times one, while  the perimeter
is  equal to the number  of picture points.  Notice that A/P = 1.070 in
Figure 16(a)  and 2.292  in Figure 16(b) for the top particle.   In Figure 16(a)
the ratio A/P = 2.877 for the bottom particle while A/P = 1.002  in Figure
16(b).  Through a number of studies we concluded that this ratio is an  excel-
lent parameter for  finding focus.

     The most logical approach appears to be  to program the calculator  to
search for minimum  A/P  and to record the data at that value of x position of
a  traverse.

SAMPLE DATA FROM HOLOGRAMS

     The primary objective of this study was  to determine feasibility for
automating data reduction.  Most of the effort was directed to the under-
standing of the properties of an individual image.  In the ultimate applica-
tion, a primary interest is in the overall particle field;  namely, statis-
tical samples, distributions and descriptions of the overall field.

                                       35

-------
     While this study was in progress , a large number of holograms  of  par-
ticle fields were produced, not only in support of this project but also  in
support of three other projects, the task of which was to provide diagnostics
of particle fields.  One of the projects involved the use of holography to
analyze fuel dispersions in explosive clouds.  The third involved a study of
combustion of fuel particles.

     In this section we present sample data from those projects to  illustrate
the ultimate capabilities of holography in this type of study as well  as  to
illustrate current limitations.  In one of these studies six holograms of a
fuel droplet cloud were analyzed in great detail.  Many hours were  spent  in
manual analysis, counting and sizing thousands of particles and studying
their distribution in space.  In the other holograms the study was  limited to
a statistical count.

     Commonly a hologram contains the images of thousands of particles, more
than enough for a statistical sample.  The number of counts required to ob-
tain a representative sample depends upon (1) the type of distribution,
(2) the accuracy of each measurement, (3) the required accuracy of  the dis-
tribution and the mean, and (4) a priori knowledge about the particle  field.
Although a detailed discussion of the statistics of particle measurement  is
beyond the scope of this report, a few approximate relations assist with  the
discussion here.

     Consider a Gaussian distribution in diameter (a log normal distribution
would perhaps be more true to nature).  If, for example, we desire  to  deter-_
mine the mean diameter we must make n measurements.  Then our mean  diameter D
has a 68 percent chance of lying within + S  of the true value, 95  percent
chance within + 2S  where S , the "standard error of the mean," is
              —   mm                                 '
approximately


                                                                           (2)
and S is the standard deviation of the n measurements about the mean.

     If ,_for example, we examine a particle field with a Gaussian distribu-
tion of D = 25 microns, and a standard deviation of 10 microns, the probabil
ity is 68 percent that error of the mean will be less than + 10 percent if
17 measurements are made to compute the mean.  To improve the probability to
the 95 percent level requires 65 measurements.  In actuality, the required
number of measurements to be made can only be determined after enough meas-
urements have been made to allow reasonable estimates of S.

     In a given recorded volume, it is normally not necessary to analyze the
entire volume.  Only enough counts to provide an accurate representative
sample are required.  And this can be determined through statistics similar
to that described above.  In each z-scan, a known volume is swept through
providing a total number and size distribution for that volume.  This pro-
duces a number density.  A second z-scan produces another set of data like


                                      36

-------
the first.  The number of  scans  required can again be  determined  after  a  few
scans are made by applying the above equation.

     The xy positions are  normally picked so that  a significant space sepa-
rates the volumes and no chance  for overlap exists.  However, when number
densities are very  low, it is advantageous to count every  single  particle in
the hologram.  In fact, more than one hologram could be  required  in  the
extreme.

     Figure 17(a)-(c) illustrates representative particle  size distributions
taken from holograms produced in this study.  Figure 17(a) was taken from a
sample of nominal 30 micron diameter glass beads.   A comparative  count made
with a microscope is  shown with  this case.  Figure 17(b) is  the analysis of
an aerosol generator, while Figure 17(c) is the combined analysis of three
holograms made on a windy  day in the Mojave Desert.
                                       37

-------
                                 Nominal 30 Micron Diameter
                                 Glass Bead Sample
                                                Microscope Sizing
                                                Holographic Recording
                                                of a Dispersed Sample
26   27   28   29   30   31  32   33   34   35   36   37   38
                    Diameter — Microns
 Figure 17(a).   Holographic size  analysis  -  glass  beads.
                            38

-------
                                    Average Number Density  530/orf
  10   20   30   40   50  60   70  80   90  100  110  120 130
                   Diameter ~ Microns
Figure 17(b).  Holographic size analysis - aerosol spray.
                            39

-------
                                  Average Number Density 150/cm~
75-
70-
65-
Cf\ _
oU ••
55-
50-
lu 45-
8 40-
M
| 35-
on _
30 *
O c _
iO
20-
i r* 	
15"
10
5-

r
^
••
••
W
^^^^^ — BB1111111
•
w

•
•r
•t
•
••
^





























































































1 1 1
1 1 1
 10   20   30   40   50   60  70   80   90  100  110 120  130
                 Diameter ~ Microns
Figure 17(c).  Holographic size analysis - dust  storm.
                           40

-------
                                   SECTION 5

                 THE  SDL FIELD-PORTABLE PARTICLE HOLOCAMERA

     SDL has developed  and has  refined during the past three  years  a  field-
portable particle holocamera for in situ diagnostics  of particle  fields in
the size range  from a few microns up to hundreds of microns in  diameter.
This system has undergone a wide variety of  field tests during  which  time its
refinement and modification to  simplify its  operation and  to  improve  its re-
liability have  continued.   Figure 18(a)-(c)  shows the system  and  a  field
installation.   The  heart of the system is a  pulsed ruby laser which delivers
a single or double  pulse of light having a time  duration of approximately 15
nanoseconds and a total energy  output to about 100 millijoules.   In the
double-pulse modej  two  such pulses can be produced with time  separations
ranging from a  few  microseconds up to approximately one millisecond.  A
helium-neon laser is  built into this unit and passes  precisely  along  the line
of the ruby laser beam  for the  purposes of alignment  and alignment  check.
The ruby laser  beam is  approximately two millimeters  in diameter when passing
from the laser.  The  first element it hits is the beam splitter which splits
off a portion of the  light to serve as a reference wave.   The remaining por-
tion of light passes  through the beam splitter to a second beam splitter
where a small portion is taken  from the laser beam for the purposes of power
output measurement.   After this, the remaining light  beam  passes  into a beam
expander whereupon  it is expanded to a two-inch  diameter collimating  lens
which collimates the  light and  passes it out through  a probe  to a 90° turning
mirror whereupon the  laser beam passes through a shuttering system  across the
sample volume.

     The purpose of the shutter system is primarily to keep the optics clean.
We have tried many  methods to keep the optics free of dust and  contamination
and we can find no  method better than keeping the optics covered  except for
a fraction of a second  during which the laser pulse exists.   The  laser is
controlled to fire  when the two protective shutters,  labeled  9  in Figure 18(a),
are in their full open  position.  The sample volume is approximately  two
inches by six inches  deep.  After the beam passes through  the sample  volume,
the scattered and transmitted light are turned 90° by a turning mirror back
into the collection system.

     The collection system consists of an image  transferring  lens system
which transfers the image of the sample volume back to a position near where
the hologram is made.

     The reference  beam passes  from the beam splitter just after  the  ruby
laser to a system of  mirrors after which it  is beam expanded  and  collimated
and then turned to  mix  with the object wave  at the hologram.


                                       41

-------
N>
       (£)  Mirror                    (§)
       (2)  Beamsplitter               (?)
       (5)  Detector                  ©
       (4)  Beam Expander/Spatial Filter©
       (D  Alignment Laser  (HeNe)   ©
Ruby  Laser
Film Plane Shutter
Film/Plate  Holder
Shutter
Port for Auto Collimator
External  Laser Alignment
Viewing Port
Sample Volume
Purged/Protective  Housing
                              Figure 18(a).   Field-portable particle holocamera.

-------


Figure 18(b)
                                      wwK^W.-iwwwffiss^Hjgojjmjj
5ide view of the holocamera installed behind a
shield as applied in explosion diagnostics.
  Figure 18(c).  Front view of  the  shielded
                             optics package.


-------
     With this system, either in-line holography or off-axis holography  is
possible.  Simply removing the reference wave converts this to an in-line
holocamera.  We have found in a number of cases that holograms made in this
system with the reference waves can be used in either mode, in-line or off-
axis, and sometimes there are advantages to being able to do both.

     The basic configuration of this holocamera is not especially novel.
Nevertheless, a number of design features built into this system were found
important in its utility in a field-type environment.  The entire system is
housed in an aluminum sealed box.  The box is purged with dry nitrogen to
keep contamination outside.  A cover lifts off for access to the inside.
A small hatch allows access to the film plane for exchanging photographic
plates.  The entire optical system inside the rectangular portion of the box
is mounted on a secondary reinforced surface which is vibration isolated by
a rubber mounting from the outside parts of the box.  This allows a certain
amount of torquing of the outside of the box without misaligning the internal
optics.

     The alignment of the ruby laser can be done without removing the cover
of the system.  This is done by inserting the optical alignment tool through
the sides of the box and alignment adjustments to the ruby laser are extended
by flexible cables to the external part of the box.  The actual condition of
alignment can be observed by peering through peep holes at the rear of the
box to the frosted glass on the back of the hologram holder observing the
reference wave and the object wave as produced with a helium-neon laser.   As
a final check, the ruby laser itself can be fired and the actual beams which
will be making the hologram can be observed on the frosted glass at the back
of the plate holder through this porthole labeled 12.

     The packaged optical system weighs approximately 100 pounds.  The power
supply which controls and powers the ruby laser is in a separate container.
It weighs approximately 200 pounds.  This entire system fits in a small
portable trailer which is used to transport the system to various places for
our application.

     We found that the holograms produced in this system in most naturally
occurring environments were of sufficient quality to be reducible by the
Quantimet image analyzer.  This is not to say, of course, that the problem
of automatic data reduction is completely solved.  A considerable amount of
work is left in automating the features which we have described which are
necessary to make the image analyzer perform the data reduction task.

     The holograms which are described in the previous section which were
ultimately used to perform the studies with the Quantimet system were made
in this system with the reference beam removed.  We were not able within the
scope of this study to do a similar analysis for the off-axis case.
                                      44

-------
                                  SECTION 6

            A PROPOSED AUTOMATIC HOLOGRAPHY DATA REDUCTION SYSTEM

     Based upon  these results,  we are now in a position to design an auto-
matic holography data reduction system.   The purpose of this section is to
present such a conceptual design based on the use of the Quantimet 720 image
analyzer currently  in operation at the EPA facility.  We will assume that  the
holograms for which this  system is designed are at least of the quality of
those tested during this  program.

     Figure 19 is a schematic layout of a suggested reconstruction system,
while Figure 20  is  a schematic layout of the electronic and control system.
The hologram would  be mounted in a computer controlled xyz traverse.   It
would be illuminated by a collimated helium-neon laser, and the reconstructed
image would then be reimaged and magnified onto the TV camera.   Therefore,
the proper positioning of the hologram in its x, y, and z dimensions would
place the three-dimensional image onto the vidicon surface.   At any instant
the image analyzer  can be addressed to provide the information  describing
the intensity distribution on the surface of the TV screen.   At any one
instant this information  includes those parameters which were mentioned in
Section 4 for images for  both in focus and out of focus and,  to a  certain
extent, noise on the TV camera.  For any one given set of holograms of a
particular quality  an operator must intervene at the beginning  to  adjust the
image analyzer such that  its adjustment falls within the optimum operating
range dependent  upon the  quality of the reconstructed image.  Such adjustment
could, in fact,  be  computer controlled;  although we believe at  this stage
computer control of everything in this process is not necessary nor desirable.
We feel that a certain amount of operator judgment will always  be  necessary
in this data reduction just as it is in almost any type of data reduction of
field data.

     The operator initially sets the thresholds on the TV screen such that a
given threshold  is  well above the noise level and, at the same  time, consid-
erably below the level of intensity of the smallest particles which are
desired to be analyzed.   This defines the particle and the display concerning
the x-y position of the particle.   At this point, sizes of those particles
which are in focus  or nearly in focus would normally be printed out as data.
A third dimension,  however, must be located to insure that only particles in
focus are analyzed  and retained as data.

     As has been described, this can be accomplished in a number of ways with
existing data outputs from the Quantimet 720.   The experiments  performed
here suggest that the two-level detection method would be a principal method
in performing this  task.   The computer which drives the hologram in its xyz


                                      45

-------
                                                                                                      Collimator
                                                                 HeNe  Laser
                               Plate or
                              Film Holder
                        High
                     Resolution
                       Camera
To Image
Analyzer
                                                                                                             To
                                                                                                          Controller
                                                                                   Traverse
                                  Figure  19.   Holocamera reconstruction system.

-------
       z-Scan
    Medium Speed
     Store Data,
     Compute A/P
                                            Preliminary Setup

                                                Set xyz limits.
                                                Calibrate for size.
                                                Set Thresholds.
                                                Set z scan speed
                                                  controls.
Figure  20.   Flow  diagram of automated data  reduction.

                           47

-------
position would be set to position the hologram at a given x-y  coordinate  and
then start the translation in the z dimension.  The annulus defined by  the
two levels of detection in this method would then be continuously  analyzed  in
the computer with reference to its ratio of area to perimeter.  When  this
ratio becomes minimum, the particle would be designated as an  in-focus  par-
ticle and the information describing that image at that time would be accepted
as data.  This would include its xyz positions, its particle dimensions,  and
the parameter area over perimeter.

     The z scan of the hologram can be carried out at a fairly high rate
until one reaches the proximity of a particle.  When the holographic  image  is
far removed from focus, no image intensity is detected on the TV screen,  and
at any such time the z scan can be run at a fairly high rate.  When the first
level of detection is surpassed by an intensity distribution,  this is an
indicator that a particle is approaching its focus.  The scan speed could at
that point be reduced.  When the second level of threshold is surpassed by
the intensity, the focus position is even more closely being reached.   At
this point, the computer-driven traverse could be slowed even further.  The
ratio of area over perimeter of this annulus would be continuously analyzed
by the computer in search for a minimum.  All data in this region of  the  scan
would be stored in a buffer memory until the focus position has been  determined.
When the ratio of area over perimeter is minimized, that data associated  with
that position would be accepted as retained data and would be transferred to a
recording disc or tape.  The remaining data in the buffer memory could, at
that point, be erased or simply renewed.  Since the xyz position of all par-
ticles is recorded, the computer can be programmed not to accept two particles
in the same xyz position such that an overlapping of the volume being scanned
could be rejected.

     When the end of the holographic volume was reached, the computer would
drive the traverse in an x dimension until a new region on the hologram was
examined once again in its z position.  It appears that a hologram of the size
which we are describing here could be reduced in this fashion in a matter of
time ranging from a few minutes to one hour, depending upon how many particles
were in the recording.

     A number of other program operations should be included to correct for
possible discrepancies, some of which have been described in Section 4.    For
example, when very large particles exist in a volume, the diffraction rings
around these particles are often counted as particles.  We have shown that  one
can reject this noise to some extent by setting resolution of the system  to  a
value at approximately equal to the smallest particle of interest.  Another
method which we believe would be required in this study is to have a volume
predetermined which would be designated as the smallest allowed volume  to con-
tain more than one particle.  If such a volume was designated, then any such
volume containing more than one particle would be rejected by the computer.
We have found in these studies that noise is characterized typically by this
kind of information.  Where noise does exist in a reconstructed image,  it
appears as a number of particles clumped in a small volume.

     The very reason for having performed this study has actually prevented
us from getting certain statistical information which will be necessary to

                                      48

-------
refine such a system.  One needs, after the construction of an automated holo-
graphic data reduction system,  to perform analysis on statistical samples of
holograms such that  factors  can be  fine tuned on some statistical basis.  Our
conclusions so far have been strictly based on  specific small number samples.
The computer must be programmed to  set certain  limits on holographically re-
constructed images which  are based  upon statistics derived from performing
studies with the system itself.  The initial setup of such an automated sys-
tem would depend upon the quality of the hologram being reduced.  However, it
is reasonable to believe  that a given set of holograms made under similar
conditions would retain the  same setting as that for the initial hologram.
We know, however, this is true only to the extent that the particle field be-
ing diagnosed is similar. The ultimate noise in the reconstructed image is
not only dependent upon the  hologram quality, but it is also dependent upon
the numbers and sizes and distributions of particles in the field of interest.
                                      49

-------
                                 REFERENCES

1.    Champagne, E.  Non Paraxial Imaging, Magnification, and Aberration-
     Properties of Holography.  J. Opt.  Soc.  Amer. 57 (1957).

2.    Trolinger, J. D.  Laser Instrumentation for Flow Diagnostics.
     AGARDograph 186 (1974).  121 pp.

3.    Trolinger, J. D.  Particle Field Holography.   Optical Engineering 5,
     p. 383 (1975).

4.    Bexon, R., Bishop, G.  D., and Gibbs, J.   Aerosol Sizing by Holography
     Using the Quantimet.   A Cambridge Instrument  Company Publication.
                                     50

-------
                                  TECHNICAL REPORT DATA
                           (Please read Instructions on the reverse before completing)
 REPORT NO.
EPA-600/2-79-005
2.
                              3. RECIPIENT'S ACCESSION NO.
 TITLE AND SUBTITLE
 EXAMINATION OF AUTOMATIC DATA REDUCTION METHODS  FOR
 PARTICLE FIELD HOLOGRAMS -
                              5. REPORT DATE
                               January 1979
                                                           6. PERFORMING ORGANIZATION CODE
 A'JTHCRJS!
 J.D. Trolinger
                                                           8. PERFORMING ORGANIZATION REPORT NO.
 PERFORMING ORGANIZATION NAME AND ADDRESS
 Soectron Development Laboratories, Inc.
 3303 Harbor Boulevard
 Costa Mesa, CA  92626
                                                           10. PROGRAM ELEMENT NO.
                                  1AD71? BK-AS  (FV  77)
                              11. CONTRACT/GRANT NO:
                                                                68-02-2491
12. SPONSORING AGENCY NAME AND ADDRESS
 Environmental  Sciences Research Laboratory — RTF,  NC
 Office of Research and Development
 U.S. Environmental Protection Agency
 Research Triangle  Park. N.C. 27711	
                              13. TYPE OF REPORT AND PERIOD COVERED
                                Final       3/77-1/78
                              14. SPONSORING AGENCY CODE
                                  EPA/600/09
15. SUPPLEMENTARY NOTES
16. ABSTRACT
       Holographic recording techniques provide  one of the most powerful particle
  field diagnostic tools in existence.  A hologram can provide a frozen
  three-dimensional image of a particle field  through which detailed microscopic
  examination of individual particles is possibe.  Frequently, a particle field may
  contain many thousands of particles, and  it  becomes impractical for the human
  operator  to glean all the data of interest from  such a hologram.  For holography  to
  reach its full potential in particle diagnostics, a three-dimensional image analyzer
  is  required.
       The  purpose of this study was to examine  the feasibility of using existing
  electro-optic image analyzers to automatically analyze three-dimensional image
  fields  and to determine what modifications of  existing equipment would be  required
  to  construct such a system.
       Sample holograms as well as holograms produced in an actual field holo-camera
  were used to make the evaluations experimentally, and well-refined analytical
  descriptions of holographic images were used to  add to the understanding of system
  requirements. The study established that  existing image analyzers are capable within
  useful  practical limits of locating particle images in three-dimensions and
  measuring size and shape factors of the particle.  A plan for integrating  such
  equipment to produce a fully automated data  reduction system is presented.	
17.
                                KEY WORDS AND DOCUMENT ANALYSIS
                  DESCRIPTORS
                                              b.IDENTIFIERS/OPEN ENDED TERMS
                                              COSATI Field/Group
 *Air pollution
 *Particles
  Examination
 *Data reduction
 *Analyzers
 holography
                                                13B
                                                14B
                                                09B
                                                14E
18. DISTRIBUTION STATEMENT


 RELEASE TO PUBLIC
                 19. SECURITY CLASS (ThisReport)
                   UNCLASSIFIED
21. NO. OF PAGES

	59
                                            22. PRICE
EPA Form 2220—1 (Rev. 4-77)   PREVIOUS EDITION is OBSOLETE
                                            51

-------