Introduction to STM Image Processing I
||Introduction to STM Image Processing I
All, or any part of the following material may be downloaded and used for
personal or educational purposes as far as
no profit is gained from its use (otherwise contact the
there is a clear indication of the © author DcsabaS TM,
with a link to this www site.
All the images presented in the following were made by using the
"STM for Students"TM
developed by DÓRA, Gyula.
by DARÓCZI, Csaba Sándor
|Make everything important more visible !
|| but hide the others!
( 2500 nm × 2500 nm, covered by Gold )
Optical gratings are used in equipments that utilize the wave
nature of light, therefore the grating constant must be
comparable to the wavelength. The fine structure of the grating
is also important, because it influences the efficiency.
Optical microscopes can not resolve this fine structure, but STM.
However, we have to have a conducting surface therefore enters the
gold as covering layer.
On the previous image the bright domains correspond to higher
areas on the sample than the dark ones. This could help us a lot
in discovering different features, however, in reality it is NOT a
strict rule, that "the brighter is higher". Therefore we should
introduce an artificial (software) light shading, that makes the
image like for example the view of a coin kept in our hands.
2-dimensional representation of a 2-dimensional data block is
rather obvious. Even though, the less sometimes is more. With a
line cut we can make more precise comparision in the height
profile of the different spots of the sample.
||Line Cut II
If we compare this line cut to the previous one, we can discover
that not only the contrast is reduced, but somehow the shapes of
the two profiles are also different. 1.) The previous line
cut seems to be more regular 2.) but it is not really
sinusoidal. 3.) The second line cut is more noisy.
||Optical Grating II
( 2500 nm × 2500 nm, covered by Gold )
The initial quality of the STM images (as we receive them from
the data acquisition system) might be different. For example, this
image of the optical grating (fabricated at
MFA (former ATKI)) is much better
than the first one. We could
think that there is no need for additional image processing in
cases like this...
||Light Shading II
However, as we have seen earlier,
light shading is well applicable for more realistic representation of
the acquired image. In this case the grating looks like it were
16000 times larger discarding the missing color of the covering
gold layer. Color can be attributed to the image, as we will see it
||Line Cut III
Before this line cut was taken we have applied a slight
smooth filtering to reduce
the high-spatial-frequency noise. The line cut shows
roughly sinusoidal profile, and the upper and the lower
half of the line is NOT so asymmetrical than in the
first case. (How can
image profile of a sample assymetrical if the
sample surface profile is symmetrical?)
||Line Cut IV
Another line cut (in orthogonal direction) reveals that there is
an irregular fluctuation of the sample surface, which is related
to the grains of the sputtered Gold layer. If we compare
the interval of fluctuations (roughly 4 nm) to the previous
line cut, we can see that most of the deviation from ideal
sinusoidal shape might be due to this origin.
Even the cleanest sample surface may contain damaged or
contaminated regions. It is not always easy to determine the
source of a given feature. "Strange" regions of an image
occasionally might be of importance.
Isolated bright or dark spikes are often just consequences
of too high bias voltage applied between the sample and
the scanning tip. These features can be eliminated from
the image by median filtering, that replaces the locally
extremal points with a kind of average of the surroundings.
(Compare the rectangular area to that of the previous image!)
Sometimes it is useful to cut out some part of a given image and
enlarge it to the size of the original one. This kind of
magnification may represent a better view, however, it doesn't
"enlarge" the information contained in the original image! In other
words, this method is NOT able to improve the resolution.
||Step by Step
( 1000 nm × 1000 nm, covered by Gold )
The quality of a newly acquired STM image can be rushed down by
several factors. It maybe possible to repaire some of them, but
the order in which the processing tools should be applied is
usually important. The point is at each step to eliminate
a given problem without influencing other properties of the image.
It may happen, that the fine positioning system had not reached
its equilibrium state when the data equisition started. This
results a side effect in the image. Of course, the best
is to repeat the measuring. If we can not (or don't want to) do
that, we can make a sub-image that doesn't contain the effected
region - in our case the dark upper edge of the previous
image. The contrast of the resulting image is better.
After we have successfully removed the side effect, all the other
problems can be seen more enhanced. Usually the next useful
thing is the subtraction of the inclination plane of the
image, because this inclination is influenced by the macroscopic
position of the sample, which is rarely important in microscopic
investigations. After having tilted back the image we gain
an additional contrast enhancement.
We discovered some little bright spikes on the previous image. Now
they have been removed by the help of median filtering. With this
method not only "point deffects" can be repaired, but erroneous lines
Here You can see a new original (as-acquired) image, and the effect
of smooth OR median filtering on it, respectively.
Would You learn more about STM Image Processing?
||Last Mod: Fri, 12 Oct 2001 07:40:28 GMT