Contrast and Microscopy
22 ˇ Basics of light Microscopy & iMaging contrast and Microscopy
All cats are grey in the dark
Why are cats grey in the dark? Here, the
term contrast comes into play. Contrast
refers to the difference of intensities or
colours within an image. Details within
an image need intensity or colour differences
to be recognised from the adjacent
surroundings and overall background.
Let us first restrict our view to a greyscale
level image, like image fig. 26. While
watching the image try to estimate the
number of grey scales that you can distinguish
­ and keep that number in
mind.
Let us now compare this number with
another example ­ you enter an office
shop in order to buy a selection of card
samples in different shades of grey. All of
the cards fall down by accident and are
scattered over the floor. Now you have to
replace them in the correct grey level order
­ how much can you differentiate
now?
Surprisingly, we are only able to differentiate
approximately 50­60 grey levels
­ that means that already the 8 bit
image on your monitor with 256 grey
scales offers greater possible differentiation
than we can discriminate with our
own eyes. The card samples in various
shades of grey need to a have an approximate
difference in contrast level of about
2% in order for us to recognise them as
different. However, if we look at the
number of image areas (pixels) that represent
a discrete intensity (grey level or
pixel intensity) within an intensity distribution
of the image we can understand
and handle contrast and brightness variations
more easily. Intensity or grey level
distributions are referred to as histograms,
see box 8 for further explanation.
These histograms allow us to optimise
camera and microscope settings so that
all the intensity values that are available
within the specimen are acquired
(fig. 27). If we do not study the available
intensity values at the initial stage before
image acquisition, they are archived and
can not be visualised in additional
processing steps.
The familiar view ­ brightfield contrast
In brightfield transmitted microscopy the
contrast of the specimen is mainly produced
by the different absorption levels
of light, either due to staining or by pigments
that are specimen inherent (amplitude
objects). With a histological specimen
for example, the staining procedure
itself can vary the contrast levels that are
available for imaging (fig. 27). Nevertheless,
the choice of appropriate optical
equipment and correct illumination settings
is vital for the best contrast.
During our explanation about Köhler
alignment (see box 2) we described that
at the end of this procedure the aperture
stop should be closed to approximately
80% of the numerical aperture (NA) of
the objective. This setting is to achieve
the best contrast setting for our eyes.
Further reduction of the aperture stop
will introduce an artificial appearance
and low resolution to the image.
For documentation purposes however,
the aperture stop can be set to the same
level as the NA of the objective ­ because
the camera sensors in use are capable to
handle much more contrast levels than
our eyes. For specimens lacking natural
differences in internal absorption of light,
like living cells (phase objects; fig. 28) or
Fig. 26: Vitamin C crystals observed in polarised
light.
Fig. 27: Histological staining. Cells obtained after a transbronchial needle application. Image and
histogram show that the overall intensity and contrast have been optimised.
contrast and Microscopy Basics of light Microscopy & iMaging ˇ 23
reflected microscopy specimens without
significant three dimensional relief structures,
the acquired image has flat contrast.
To better visualise the existing image
features (fig. 28, original image),
subsequent digital contrast optimisation
procedures may be applied (fig. 28, improved
image). But to visualise more details
in those specimens optical contrast
methods must be used.
Like stars in the sky ­ Darkfield Contrast
Dust in the air is easily visible when a
light beam is travelling through the air in
a darkened room. The visibility is only
achieved because the dust particles diffract
and/or reflect the light and this light
is now travelling in all directions.
Therefore we can see light originating
from the particle in front of a dark background
even when the particle itself is
too small to be resolved or does not show
an appropriate contrast under daylight
conditions. This phenomenon is also used
in darkfield (or dark ground) microscopy.
Light is directed to the specimen in a way
that no direct light enters the objective. If
there is no light scattering particle the
image is dark, if there is something that
diffracts or reflects the light, those scattered
beams can enter the objective and
are visible as bright white structures on
a black background (fig. 29).
(See also: http://www.olympusmicro.com/
primer/techniques/darkfieldindex.html)
Transmitted darkfield
For transmitted microscopy including
stereo microscopy, this contrast method
is especially used to visualise scattering
objects like small fresh water micro-organisms
or diatoms and fibres (fig. 30).
Almost all upright microscopes can be
easily equipped for darkfield illumina-
tion.
The easiest way to achieve simple
darkfield is a central light stop insert for
the condenser. For better illumination or
even high resolving darkfield, special
darkfield condensers are required
(fig. 29).
To ensure that no direct light is entering
the objective, the numerical aperture
(NA) of the condenser has to be about
15% higher than the NA of the objective.
This is in contradiction to all other contrast
methods where the objective has a
higher or equal NA than the condenser.
Remember the NA is a number that describes
the angle of the light cone a condenser
or objective is using. An objective
with high NA like the apochromate is
characterised by a high angle of its light
cone and therefore it may be possible that
some of the direct illumination will also
enter the objective. This would destroy
the darkfield contrast immediately. For
that reason objectives with high NA are
Fig. 28: Brightfield image of
living mouth epithelial cells on a
slide before and after digital
contrast optimisation of the
archived image and corresponding
histograms. See the section
"Making it look better", for
further explanation.
Box 8: Histogram optimisation during acquisition
An intensity histogram depicts the intensity distribution of the pixels in an image. The intensity values are
plotted along the x axis and range from 0­255 in an 8-bit greyscale image or 24-bit (3x8bit) colour image.
The number of pixels per intensity value is displayed along the y axis.
The intensity histogram provides the means to monitor general characteristics of a digital image like its overall
intensity, its contrast, the dynamic range used, possible saturation, the sample's phases etc. In this respect
the histogram gives a more objective criterion to estimate the quality of an image than just viewing it (which
is somewhat subjective).When displayed in the live mode during image acquisition, the histogram allows optimising
and fine tuning of the acquisition modes and parameters on the microscope as well as the camera.
These include microscope alignment, contrast method, microscope settings, or current camera exposure time.
So the histogram helps to acquire a better image containing more image information.
Left figure is obviously lit correctly but has poor contrast. The corresponding histogram mirrors this: the peak
is roughly located in the middle of the intensity range (x-axis), so the camera's exposure time is correctly set.
But all pixels are squashed in the middle range between about 75 and 200. Thus, only about half of the dynamic
range of the camera system is used. Aligning the microscope better (e.g., light intensity) and adapting
the camera exposure time increases the image contrast (right figure). It also spreads the intensity distribution
over the whole dynamic range without reducing information by saturation (see the corresponding histogram).
Here, more detailed image structures become visible.
General rule
The overall contrast is usually best when the intensity histogram covers the whole dynamic range of the system;
but one should usually avoid creating overflow or saturation at the right side of the histogram, i.e. white
pixels.
It is also possible to use the histogram to improve the image contrast afterwards. However, you can not increase
the image content; you can only improve the image display.
Epithel cells viewed with phase contrast: left side with low contrast setting, right sight with
contrast optimisation of microscope and camera setting.
24 ˇ Basics of light Microscopy & iMaging contrast and Microscopy
Fig. 29: Light path for darkfield compared to brightfield set up in transmitted and reflected illumination.
available, designed with an internal iris
diaphragm to reduce the NA to the appropriate
amount for darkfield observation.
Reflected darkfield
Within the reflected microscopy applications
the darkfield illumination is a very
common contrast technique. It allows the
visualisation of smallest scratches and
changes in height because of the circular
oblique illumination (fig. 31). To achieve
a reflected darkfield, the amount of special
adaptations at the microscope is
more sophisticated. The central light stop
is located within a cube of the reflected
light attachment and the special brightfield/darkfield
objective guides the illumination
light in an outer ring to the
specimen. Only the scattered light from
the specimen runs in the normal central
part of the objective as image forming
light rays (fig. 29). Those objectives can
also be used for normal brightfield observation
or other techniques like differential
interference contrast (DIC) and/or
polarisation.
Creating destructive interference Phase
Contrast
Light that is travelling through part of a
specimen and is not absorbed by amplitude
objects will not produce a clearly
visible image. The intensity remains the
same, but the phase is changed compared
to the light just travelling in the
surrounding areas. This phase shift of
about a quarter wavelength for a cultured
cell is not visible to our eyes. Therefore,
additional optical elements are
needed to convert this difference into an
intensity shift. These optical elements
create a contrast where un-deviated and
deviated light are  wavelength out of
phase, which results in destructive interference.
This means that details of the
cell appear dark against a lighter background
in positive phase contrast (see
figures in box 8). (See also: www.olym-
pusmicro.com/primer/techniques/phase-
contrast/phaseindex.html)
For phase contrast microscopy two elements
are needed. One is a ring slit insert
for the condenser, the other is special
objectives that contain a phase plate.
Objectives for phase contrast are characterised
by green lettering and an indication
of the size of the phase ring like Ph1,
Ph2, Ph3 or PhC, PhL, PhP. Corresponding
to these objective phase rings the appropriate
inserts for the condenser have
to be used.
Both elements are located at the so
called back focal planes. They are visible
Box 10: Alignment of phase contrast
1. A microscope condenser with Köhler illumination has to be in Köhler positioning.
2. Remove one of the eyepieces and look into the empty eyepiece sleeve.When using a centring telescope at
the eyepiece sleeve (highly recommended and in some inverted microscopes already build in at the tube
e.g. Olympus U-BI90CT ­ binocular tube), bring the bright ring (condenser ring slit) and dark ring (objective
phase plate) into focus by turning the eye lens focus.
3. Use the centring screws for the condenser inserts
to centre the phase contrast ring, so that
the bright ring overlaps the dark ring within
the field of view (see figure).
4. Repeat these steps for each phase and contrast
ring set.
5. Remove the centring telescope and replace it
with the eyepiece.
6. Widen the field iris diaphragm opening until the diaphragm image circumscribes the field of view.
Box 9: Alignment of a transmitted
darkfield condenser
1. Engage the 10x objective and bring the specimen
into focus.
2. While looking through the eyepieces and using
the condenser height adjustment knob, carefully
adjust the height of the condenser until a
dark circular spot becomes visible (Figure A).
3. Turn the condenser centring screws to move
the dark spot to the centre of field of view
(Figure B).This completes the centration.
4. Engage the desired objective. Using the condenser
height adjustment knob, adjust until
the darkfield spot is eliminated and a good
darkfield image is obtained.
contrast and Microscopy Basics of light Microscopy & iMaging ˇ 25
when an eyepiece is removed and can
then be aligned under optical control (for
better viewing a centring telescope
should be used).
Due to the optical principles of phase
contrast it allows good contrast in transmitted
light when the living specimens
are unstained and thin. A specimen
should not be more than 10 m thick. For
these specimens the dark contrast is
valid, whereas thicker details and overlaying
structures produce a bright halo
ring. This halo-effect can be so strong
and superimposed that a clear analysis of
the underlying morphology becomes critical.
Nevertheless, the artificial looking
image of thicker structures can be used
by expert eyes to determine how many
cells within an adherent cell culture are
undergoing mitosis or have entered cell
death pathways. Fig. 32 shows bright
halo rings, visible round the relatively
thick astrocyte cell bodies, whereas the
fine details show dark phase contrast.
When observing cells in a chamber
e.g. those of a 24 well plate, the cells
which are best for imaging may lie at the
border of the well. It is possible to see the
cells, but because the light path is
changed by the border of the well and
the adhesion effect of the medium, the
phase contrast is totally misaligned at
this position. We can realign the phase
rings for this specific specimen position
as described below, but will have to realign
them again when the stage is moved
to other positions. Furthermore, the use
of the special Olympus PHC phase contrast
inserts instead of the PH1 helps to
ensure better contrast in multi well
plates where meniscus problems are ap-
parent.
Making it look better ­ for vision only
It is essential to select the optical contrast
method appropriate to the sample
investigated and depending on the features
which are to be made visible. You
can only build upon something which is
already there.
Having done the best possible job here
you can try to better visualise these features.
It often makes sense to adjust the
image contrast digitally. Here you can either
elevate the overall image contrast or
you can emphasise special structures.
Whatever you want to do, you should
again utilise the histogram to define exactly
the steps for image improvement.
Fig. 28 shows a drastic example (left
side). The optical contrast is so poor that
the cells' relief is hardly visible. The corresponding
histogram shows that the
Fig. 32: Astrocytes with phase contrast, note: cell
bodies are surrounded by halo rings.
Fig. 33: Epithelial cells scratched with a spoon from the tongue and transferred
to a cover slip. A,B: DIC illumination at different focus plane; C: Oblique contrast;
D: Olympus Relief Contrast imaged with lower numerical aperture.
Fig. 30: Epithelial cells with transmitted darkfield
contrast.
Fig. 31: Wafer at reflected darkfield contrast.
26 ˇ Basics of light Microscopy & iMaging contrast and Microscopy
Fig. 34: Simple principle of Nomarski DIC microscopy.
pixels are grouped in the centre of the
histogram indicating that only a small
fraction of the camera's dynamic range
has been used (intensity values 165­186
out of 256). The image has only 21 different
intensity values. The present features
can be made visible by stretching the histogram
over the whole dynamic range
from 0­255. This operation does not
change the information content of the
image but lets us at least see what is
there (right side).
This stretching comes to its limits as
soon as there is a single black and a single
white pixel in the image. In this case
a moderate cut of up to 3% on the left
side (dark pixels) and on the right side
(bright pixels) doesn't cause much information
loss but increases the contrast of
the image's main features in the central
intensity range. To selectively accentuate
features of a distinct intensity range (socalled
phase), an individual transfer
function has to be defined which increases
the phase's contrast to the background
or other phases. The intensity
and contrast image operations are usually
performed after image acquisition.
Often, the camera control of the image
analysis software makes it possible to
have the contrast stretched image calculated
and displayed in the live mode.
Alternatively, the contrast can be set
manually on the live histogram while the
direct results can be monitored in the
live image. Whatever way the digital
histogram is used, it is a powerful tool
to maintain image fidelity as well as
to create a clear picture of physical na-
ture.
Light alone is not enough ­ Differential
Interference Contrast (DIC)
If you have ever been skiing on a foggy
day, you will know that diffuse light
strongly reduces the ability to see differences
in the height of the snow ­ and that
can cause major problems. An equally
distributed light source does not produce
clear shadows and this causes reduced
visibility of three dimensional structures.
Our human vision is triggered to see
three dimensions and is well trained to
interpret structures if they are illuminated
more or less from one point. The
resulting dark and bright areas at the
surface of a structure allow us to easily
recognise and identify them. Using our
experience, we get information of height
and distance. Therefore, a contrast
method that displays differences in a
structure as a pattern of bright and dark
areas is something that looks very familiar
to us and seems to be easy to interpret.
Structures within a specimen can
be identified and even though they are
only two dimensionally displayed they
look three dimensional.
Real three dimensional images can
only be observed at a stereo microscope
where two light paths of two combined
microscopes are used, sending the image
to our eyes at a slightly different angle.
But this will be a topic to be described
later.
Simple from One Side
The simplest way to achieve a contrast
method that results in a relief-like image
is by using oblique illumination: however,
the theory of oblique illumination is more
complex. Details can be found at: www.olympusmicro.com/primer/
techniques/oblique/obliquehome.html.
Special condensers
are available to handle this technique
on upright microscopes with a lot of comfort.
Oblique illumination is also often used
on stereo microscopes to enhance the contrast
of a 3D surface simply by illuminating
the specimen from one side. For reflected
light, this can be done with flexible
cold light fibres, or with LED ring light systems
that offer reproducible illumination
of segments that can even be rotated.
Box 11: How to set up the transmitted DIC microscope (e.g. Olympus type)
J Use proper Köhler positioning of the microscope with a specimen in focus (best with 10x objective).
J Insert polariser (condenser side) and analyser (objective side) into the light path.
J Remove specimen out of the light path.
J Turn the polariser until the image gets to its darkest view (Cross-Nicol position).
J Insert the DIC prism at the condenser corresponding to the objective and slider type in use, immediately
the image is bright again.
J Insert the DIC slider at the objective side. By rotating the fine adjustment knob at this slider the amount of
contrast of the DIC can be varied.
J The best contrast is achieved when the background shows a grey colour and the specimen is clearly
pseudo three dimensional.
Another Way to Achieve this Setting Is the Following:
J Use proper Köhler positioning of the microscope with a specimen in focus (best with 10x objective).
J Insert polariser (condenser side) and analyser (objective side) and the prism slider into the light path.
J Remove specimen out of the light path.
J Remove one eyepiece and if available look through a centring telescope, rotate the slider
adjustment knob until a black interference line is visible (box 11- Figure ).
J Rotate the polariser until the black line becomes darkest.
J Insert the DIC prism at the condenser and insert the eyepiece for normal observation.
J Fine adjustment of DIC-slider can be done for selecting the amount of interference contrast.
contrast and Microscopy Basics of light Microscopy & iMaging ˇ 27
In transmitted microscopy the oblique
condenser (e.g. the Olympus oblique condenser
WI-OBCD) has an adjustable slit
that can be rotated. After Köhler alignment
of the microscope this slit is inserted
in the light path and results in illumination
from one side so that
specimens that vary in thickness and
density are contrasted. Rotating the slit
enables us to highlight structures from
every side. The contrast itself is produced
by the complete thickness of the specimen
and the resolution of the image is
limited due to the oblique illumination
(fig. 33c).
To overcome the limitations of oblique
contrast, Nomarski Differential Interference
Contrast (DIC) is commonly used for
high resolving images. The benefit of this
method is that the relief like image is
only contrasted at the focus area (depth
of field). The user can optically section a
thicker specimen by changing the focus
level.
As shown in fig. 33 the contrasted focus
layer can be restricted to the layer of
the somata (fig. 33a) or the superficial
part of these cells (fig. 33b). In addition,
using infrared light (mostly used around
700 or 900 nm) instead of white light,
this technique allows a very deep look of
more than 100 m into thick sections,
which is often used in neurobiological research.
Nomarski DIC creates an amplified
contrast of phase differences which
occurs when light passes through material
with different refractive indices. Detailed
information about the theory and
use of the DIC method can be found at
www.olympusmicro.com/primer/techniques/dic/dichome.html.
Here we will
concentrate on the basic introduction
and the practical use.
To achieve transmitted Nomarski DIC
images, four optical elements are needed:
a polariser, two prisms and an analyser.
For the reflected DIC setup only one DIC
prism (the slider) is required.
Let us have a closer look at transmitted
DIC. The wave vibration direction of
light is unified by a polariser located between
the light source and condenser
Fig. 37: Transmitted polarising microscopy; variation of melted chemicals viewed with crossed polariser.
Images courtesy of Norbert Junker, Olympus Europa GmbH, Hamburg, Germany.
Fig. 35: Specimens imaged with different shearing values. Left side shows thin NG108 cells imaged
with the high contrast prism set, the middle image shows a thick diatom specimen imaged with the
high resolution prism set and the right image shows PtK2 cells imaged with the general prism set.
Fig. 36: Alignment of Olympus Relief Contrast or
Hoffmann Modulation condenser insert, grey
and dark areas are located within the objective
and areas A and B are located as an insert within
the condenser and can be aligned accordingly.
28 ˇ Basics of light Microscopy & iMaging contrast and Microscopy
(fig. 34). In the condenser, a special insert
­ a Wollaston prism (matching the
magnification of the objective) ­ divides
every light ray into two, called the ordinary
and the extraordinary, which vibrate
at a 90 degree angle to each other.
Both light rays then travel a small distance
apart, the so called shearing distance.
At the specimen, the ray that is
passing through e.g. a cell part is delayed
compared to the one passing through the
surrounding medium. This result in a
phase shift of both rays, which are recombined
with the help of a second prism
located at the objective revolver. Only
those combined rays with a phase shift,
interfere in a way that they contain vibration
planes that pass through the analyser.
To create an easily observable
pseudo 3D image, a prism can be moved
in and out to enhance the phase shift between
ordinary and extraordinary ray. At
a mid position the background should be
dark and the specimens should be visible
as if illuminated from the equivalent
North-West and South-West simultaneously.
When the slider is screwed more to
one direction the typical three dimensional
view of the specimen will come up,
either having the sun in North-West or
East-South location. This will only work
if no depolarising material (e.g. plastic) is
used within the light path. Therefore, a
real high resolving DIC method can not
be used with plastic Petri dishes or multi
well plates. In those cases other methods
like Hoffmann Modulation Contrast or
the analogous Olympus Relief Contrast
are commonly used.
The DIC method does not require special
objectives, but the prism has to match
the objective used. Suitable prisms are
available for most fluorite and apochromat
objectives. DIC contrast can be offered
with shearing values for every optimisation,
a high contrast set-up that is
best for very thin specimens (higher
shearing value, fig. 35 left), a standard
prism slider for general use (fig. 35 right)
and a high resolution slider for thick
specimens (lower shearing value, fig. 35
mid) like C. elegans or zebra fish em-
bryos.
Different names but the same principle:
Hoffman Modulation Contrast ­ Olympus
Relief Contrast
Whenever high resolution is needed and
plastic dishes are used, the Hoffman
Modulation or Olympus relief contrast
method are common techniques for inverted
transmission light microscopes
e.g. for in vitro techniques (fig. 33d).
In principle it combines the oblique illumination
of a specimen where the reBox
12: How to describe image processing operations mathematically
From the mathematical point of view, an image processing function is a transfer function which is applied to
each pixel of the original image individually and generates a new pixel value (intensity/colour) in the resulting
imaging. The transfer functions can be roughly divided into point operations, local operations and global op-
erations.
J Point Operations: The resulting pixel is only dependent on the intensity or colour of the original pixel, but
it is independent from the pixel places (x/y) and the pixel intensities/colours of the neighbouring pixels.
Point operations can be monitored, defined and executed via the histogram.All intensity and contrast operations
are point operations.To be exact, point operations are not filters in the more specific sense of the
word.
J Local Operations: These look not only at
the pixel itself but also at its neighbouring
pixels to calculate the new pixel
value (intensity / colour). For example,
convolution filters are local operations.
Many well known noise reduction,
sharpening and edge enhancing filters
are convolution filters. Here, the filter's
transfer function can be described as a
matrix consisting of whole positive or negative numbers, known as weight factors.Any original pixel is situated
at the centre of the matrix.The matrix is applied to the original pixel and its neighbours to calculate
the resulting pixel intensity. See box 12 ­ Figure 1a as illustration.The 3x3 matrix has nine weight factors
W1 ­ W9. Each weight factor is multiplied with the respective intensities I1 ­ I9 of the original image to
calculate the intensity of the central pixel I5 after filtering. See a concrete numerical example in box 12 Figure
1 b. In addition to this calculation, the resulting intensity value can be divided by a normalisation
factor and added to an offset value in order to ensure that the filter operation does not alter the average
image brightness. The matrix weight factors determine the function and the strength of the filter. For example,
a 3x3 matrix with all weight factors being one makes a simple average filter used for noise reduction.
Sharpen filters consist of matrices having a positive central weight factor surrounded by negative
weight factors.The unsymmetrical matrix in fig. 43b creates a pseudo topographical contrast in the resulting
image.
J Global operations: All pixels of the original image with regard to their intensity / colour and position (x/y)
are used to calculate the resulting pixel. Lowpass and bandpass noise reduction filters are examples. In
general, all Fourier Transform filters are global operations.
b)
a)
Fig. 38: Brightfield image of a wafer. Left side: The ring shaped background pattern is clearly visible.
Right side: The same image with real-time shading correction.
contrast and Microscopy Basics of light Microscopy & iMaging ˇ 29
fraction differences at various parts of
the specimen shape are used for contrast
enhancement. It allows semi-transparent
specimens structures to be analysed in a
manner difficult to achieve using bright
field microscopy. (See also
www.olympusmicro.com/primer/techniques/
hoffmanindex.html).
For this contrast technique a special
condenser and objective are needed. In
advanced systems the condenser is
equipped with a polariser and a condenser
slit plate (according to the objective
in use, fig. 36 A and B area). The
achromatic or fluorite objective contains
a modulator insert at one side of the back
focal plane (fig. 36 dark and grey area).
The slit plate at the condenser has to be
aligned according to the modulator
within the objective (fig. 36). This can be
done by visualising these elements via a
focused centring telescope (remove one
eyepiece and insert the telescope). The
resulting contrast can be varied by rotating
the polariser at the condenser.
The interpretation of DIC and relief
contrast images is not intuitive. These
techniques contrast different refractive
indices within a specimen into a pseudothree-dimensional
image. This means
that specimen details which look like
holes or hills on the surface of a structure
(see fig. 35 left and right side) may
simply be areas of different refraction index
but not necessarily different in
height.
Get more then expected ­ Polarisation
DIC equipment on a microscope allows
the user to employ a totally different microscopic
method, simple polarisation. In
this case no prisms are used and only the
first settings of aligning polariser and analyser
in crossed positioning (see box 11)
are needed. This setting will generate a
dark image because the vibration direction
of light that is travelling through the
specimen is exactly the vibration direction
that is totally blocked by the analyser.
However, if the specimen contains
material that is able to turn the light,
some light can pass the analyser and is
observed as a bright detail on a dark
background. Examples of such anisotropic
materials are crystalline Vitamin
C, emulsions of butter, skeletal muscles,
urate crystals (Gout inspection), amyloid,
rocks and minerals, as well as metal surfaces
or the DIC prism itself. The list is
very long and often the combination of
simple polarising microscopy and DIC
can offer a more detailed analysis of the
specimen, simply by using equipment
that is already available. Beside all the
analytical advantages, polarising microscopy
also offers images of high aesthetical
value (fig. 37).
Revealing structures with imaging
filter techniques
We are now at the point where the suitable
microscope contrast method has
been selected and the optimum camera
settings have been made. So the field has
been tilled and digital filter and image
processing techniques come into play.
This is explained in more detail below.
Along a thin line
Image processing can turn the photograph
of an ordinary face into an unbelievable
beauty or create images that
have no counterpart in the real world.
Knowing this, digital image processing is
often compared with manipulation of results.
It might be true for many microscopists
that they enjoy the amazing
beauty of hidden nature as being mirrored
in the images they acquire, but the
focus of the daily work lies upon the actual
information within the image. Yet
this information can be superimposed by
artefacts which might be created by
specimen preparation, illumination conditions,
camera settings or display parameters.
At this point, filter techniques
come into their own right.
Depending on the reason for imaging,
a wide variety of processing steps can be
applied to reveal new information or to
enhance image clarity for the details that
are under observation. What kind of artefacts
may appear in a digital image?
Imperfect digital images may look weak
in contrast, unevenly illuminated,
wrongly coloured, diffuse, blurred, noisy,
dirty etc. These distortions might make
the image look unprofessional. But what
is more, they might make any automatic
measurement routine based on threshold
values or edge detection difficult and
sometimes even impossible to apply.
Fig. 39: Fluorescence image of stem cells. The detail
zoom better reveals the noise level.
30 ˇ Basics of light Microscopy & iMaging contrast and Microscopy
Many of these artefacts can be reduced
by using a suitable digital filtering technique.
Often there exist several different
filters to improve the image quality. It is
not always easy to find the best method
and parameter settings to suppress the
image defects without disturbing the
"real" image information. This is the
highest standard which the optimum filter
should meet: removing the image artefacts
and keeping the specimen structures
in the image. This goal cannot
always be accomplished. When doing image
processing, there is often a thin line
between what needs to be done and what
would be good to accomplish. In addition,
there is a thin line between scientific
virtue and creative composition!
Now or later ­ both are possible
Digital image processing is usually applied
after image acquisition and image
storage. But the latest image processing
systems support "Live Digital Image
Fig. 40: The same image as in fig. 39 after application
of the Sigma noise reduction filter. The detail zoom
makes the improvement better visible.
Processing", which means that numerous
real-time functions can be executed
during image acquisition. These include
the online histogram, which is used to
monitor the image brightness and contrast.
Additionally, the histogram function
itself can be used to have the image
contrast improved automatically or to set
it in the live mode by hand. (How to use
the histogram to improve image contrast
see box 8.) Another real-time function is
the white balance, which corrects colour
shifts at the moment of acquisition. Nonuniform
specimen illumination can also
be immediately corrected in the live image
using the online shading correction.
This is described below. The operations
of contrast enhancement, white balance
and shading correction can be applied
during or after image acquisition. They
are point operations, mathematically
speaking (refer to box 12). This makes
them suitable for reducing image artefacts
without distorting the image content
itself.
Where there is light, there is also
shade
One major source of distortion in a light
microscopic image is brightness fluctuation,
which may arise from out-of-axis illumination
conditions or when the microscope
is not aligned optimally. Uneven
illumination artefacts may also appear
under regular observation conditions using
low objective magnification and/or in
combination with low magnifying camera
adapters. Then the shading artefacts
are manifested as a ring shaped shade
structure which becomes even darker
near the edges of the image (fig. 38 left
side). The minor shading within the background
needs to be corrected. The operation
applied here is called background
correction, flat-field correction or shading
correction. (A simple tool for background
correction can be downloaded at
www.mic-d.com.)
The best method is to record a background
image and a dark image in addition
to the "raw" specimen image. The
background illumination profile is simulated
in the background image, whereas
the noise level of the camera system is
apparent in the dark image. The advantage
here is that these images are gathered
independently from the specimen
image.
The background image is usually obtained
by removing the specimen and
leaving an area of mounting medium and
cover slip in place. If this is not possible,
the same result can also be achieved by
leaving the specimen in place but to defocus
the microscope. It is important that
the background image does not show any
debris. The dark image is acquired by
closing the camera shutter. It should use
the same exposure time as the specimen
image.
The shading correction algorithm first
subtracts the dark image (d) from the
specimen image (s) as well as from the
background image (b). Then it divides
the corrected background image through
the corrected specimen image: (s­d)/(bd).
This procedure can also be performed
automatically with the live image. Here,
the background and the dark images
have to be acquired only once for each
microscope objective using a standard
sample and standard acquisition parameters.
This is how the image in fig. 38
(right side) was taken.
If the background and dark images
are not available, the shading correction
can also be derived from the specimen
image itself. For example, a very strong
smoothing filter averages out all structures
and reveals the background. Or a
contrast and Microscopy Basics of light Microscopy & iMaging ˇ 31
multi-dimensional surface function can
be applied to fit the background illumination
profile.
Just too much noise
Random noise can be an annoying phenomenon
which is encountered when the
specimen signal is low and/or was enhanced
with a high gain factor. An example
from fluorescence microscopy is
shown in fig. 39. In regular brightfield
light microscopy images random noise is
usually not visible. But it is often unsheathed
when a sharpen filter is applied
to the image. See fig. 42, first and second
image as an example. There are different
smoothing filters which can be used to
reduce the noise. Smoothing filters are
also applied to suppress artefacts deriving
from small structures like dirt, dust,
debris or scratches. Yet the challenge is
to find a filter which eliminates the noise
and the artefacts without smoothing the
object edges too much.
Fig. 41 shows an artificially created
image having only one real structure: a
simple rectangle of light shading upon a
dark background. This structure is superimposed
by two different kinds of distortion:
strong statistical noise interference
on one hand and so-called hot pixels
or shot noise on the other hand. The shot
noise is individual bright and dark pixels
and can come from a defective camera.
These pixels also mirror artefacts from
dust particles, dirt, debris or small
scratches.
The noise reduction filters applied to
the image are local or neighbouring operations,
i.e., the neighbouring pixels are
taken into account to calculate the new
pixel value of the resulting image (see
also box 12). The Mean and NxN filters
just average everything and thus also
broaden the structure of the rectangle
(fig. 41). The more powerful the smoothing
effect, the more noticeable the morphology
of the object within the image
will be altered. Shot noise will not be removed.
Whereas the Mean filter applies
fixed settings, the extent to which the
NxN filter averages depends on the parameters
set.
The Sigma or Gaussian blur filter is a
special average filter which does not affect
any pixels deviating greatly from
their surrounding area (fig. 41). So "real"
structures and edges like the rectangle
are not touched whereas the random
noise disappears. The Sigma filter is the
method of choice to reduce selectively
statistical noise without greatly broadening
the specimen structures (see also example
in fig. 40).
The Median and Rank filters eliminate
the shot noise widely (fig. 41). They are
especially suited to suppress dust and
dirt, in general all small structures which
stand out the underground. Comparing
the effects the different smoothing filters
have on the artificial sample image (fig.
41), the image's two different distortions
could be eliminated best by applying two
filters successively: first the Sigma filter
against the statistical noise, and then the
Rank filter against the shot noise. This
leads to the somewhat philosophical rule
which says that it is best not to strike
everything all at once but to cut the problem
into pieces and find the appropriate
measure to deal with each piece sepa-
rately.
NxN-, Sigma- and Rank filters have
user-definable controls which make it
possible to adjust the smoothing effect
optimally to the image. There is another
option to reduce statistical noise, usually
with fixed parameters, which gives quite
good results: the lowpass filter. This filter
is a global operation (refer to box 12)
which filters out high frequency and periodic
distortions. Yet here, the edges become
somewhat broadened. An even better
global operation is the bandpass filter
which reduces the noise and preserves
the steepness of edges.
Revealing the details
Another class of filters seen from the application
point of view are the sharpen
filters. Sharpen filters can be used to enhance
fine image details. After processing
an image with a sharpen filter, the
image appears to look clearer. But
sharpen filters have to be applied somewhat
carefully because they can create
artefacts themselves if overdone. The established
sharpen filters are local neighbouring
operations like the convolution
filters (see box 12). The newer filters use
the so-called unsharp mask algorithms.
(Please have a look at fig. 42.) Sharpening
the first, original image, gives the
second image. Unfortunately, this enhances
not only the structure but also
the noise present in the original image.
So the sharpen operation makes the
noise visible. Additionally, the filter parameters
have been set too aggressively
so that the "lamellae" structures of the
diatom look artificial. This context has to
be kept in mind when applying a sharpen
filter: Before the sharpen filter can be
applied to an image, a noise reduction
filter must often be applied first. Any
Fig. 41: Comparison of different smoothing filters. See text for further explanation.
32 ˇ Basics of light Microscopy & iMaging contrast and Microscopy
sharpen filter with user controls should
be applied conservatively. Here, less is
often more. The third image in fig. 42
shows the result after noise removal via
a Sigma filter and successive moderate
sharpening.
The DCE filter is a specially designed
sharpen filter and is part of the Olympus
image processing software solutions. The
letters DCE stand for Differential Contrast
Enhancement. The DCE filter enhances
weak differences in contrast. It
selectively takes the lesser intensity modulations
between neighbouring pixels
and enhances them, while greater intensity
modulations remain as they are. So
the DCE filter renders image structures
visible which are barely distinguishable
in the original image. The filter works
the better the more fine structures the
image has. Resulting images are more
detailed and appear more focused. See
fig. 42, last image, and both images in
fig. 43 as examples.
Putting everything together
To integrate what has been said above,
there are some image processing operations
which are meaningful or even necessary
to apply to an image, either in real
time or after acquisition. These operations
can reduce image distortions, prepare
the image for automatic measurements
or just make the image look better.
The following steps suggest a possible
strategy to proceed with digital image
operations when acquiring an image (see
also:
www.olympusmicro.com/primer/digi-
talimaging/imageprocessingsteps.html)
1. Shading correction is to equalise uneven
background illumination.
2. Contrast enhancement is to optimise
brightness and contrast.
3. White balance is to adjust colour
shifts.
4. Smoothing is to reduce random noise
and suppress shot noise and small artefacts
like dust and dirt.
5. Sharpening is to enhance fine edge
detail.
Here, we have covered a few of the wide
range of possible filter operations. There
are many more which go beyond scientific
applications and transcend the pure
picturing of the visible reality towards
creative, funny, intelligent and beautiful
compositions.
Fig. 43: Epithelial cells viewed with phase contrast (left side) and transmitted darkfield contrast (right
side). The original images are acquired with optimised microscope settings. These have already been
shown in box 8 and fig. 30. The lower right parts of the images are each filtered with the DCE filter.
Fig. 42: Brightfield image of a diatom. (a): original image. (b): after application of a strong sharpen filter. This increases the noise. (c): after application of the
Sigma noise reduction filter and a moderate sharpen filter. (d): after application of the Sigma noise reduction filter and the DCE sharpen filter.
contrast and Microscopy Basics of light Microscopy & iMaging ˇ 33