B. Scaling/contrast enhancement top | MacLispix Home
The linear features are shown with low contrast. This is because the sample consists of thin gold layers in nickel which gives an inherently low contrast image, especially at lower electron beam voltages or with a slightly contaminated surface. The scaling is also thrown off by some of the image features: The default scaling for the image shows the minimum value as black and the maximum as white. The few small black dots at the top of the image and to the right of the diamond shaped indent mark throw this type of scaling off. There are not many dark pixels as seen from the histogram:
Invoke the MLx -> show values -> histogram command. The histogram
is a plot of the number of pixels with a given intensity vs. the intensity.
The up-arrow key will magnify the y-scale of the histogram plot. Pushing
that key four or five times will show the few dark pixels on the left of
More detail can be shown using histogram
equalization, which assigns equal numbers of pixels to each gray level as
far as possible. That is, equal brightness increments will correspond to
equal pixel counts. Use the MLx -> Scaling -> Equalize command, which
works with byte images (as this one is) or integer images, but not real
valued images. For equalization or other types of scaling, real valued images
can be converted to integer valued images with the MLx -> Scaling ->
Prescale command. Histogram equalization thus attempts to make the histogram
as flat as possible. (Display the histogram of the now equalized image,
again with the MLx -> show values -> histogram command. There are
lots of spikes and valleys in the histogram, as well as the dome shape because
of the constraints imposed by the limited precision of this data -> the
gray levels or pixel values consist only of the integers from 0 to 255.
This type of enhancement tends to bring out detail in the background and
to wash out detail in the smaller features, because most of the image area
is in the background. Equalization brings out unwanted detail in this case,
showing the smudges on the surface rather than the parallel lines.
Scaling operations are done on the original data, not necessarily on what you are looking at in the window. In other words, when another scaling operation is done on this window, such as the MLx -> Scaling -> % Outliers command below, the original data will again be used. If you wish to do serial scaling operations, give the scaled data a name with the MLx -> Image Windows -> Name Array - Scaled command, display that scaled array, and do the next scale command on the new image data.
A less severe type of scaling, often useful in micrographs, is to scale
the intensities linearly as before, except to discard the brightest and
darkest pixels. Using the MLx -> Scaling -> % Outliers command, type
in .5 or 0.5 for the percentage of pixels to discard. The result is a useful
amount of contrast enhancement (if the image window is not in front, click
on any part of it that is visible to bring it there, or select it with the
Windows menu or by holding down the option key and using the W menu.). The
brightest and darkest valued pixels are clipped white and black respectively,
causing loss of detail in the small areas (1/200 of the image in this case)
associated with them. Here, the effect is seen as a deepening of the shadow
on the dark side of the indent mark, and a brightening of the highlights
on the bright side. This does not affect our analysis, as we are interested
in the calibration lines, not in the surface of the indent mark.
(Optional) Alternatively, the contrast can be changed manually by sliding the black and white limits on the contrast slider. Click on the SRM484 window to make it the front window, if it is not already. The type in the title bar of the front window is black, and is gray for all other windows. Then, select the MLx -> Threshold Slider command. Control -> double click on the slider until the bar appears gray. Click and hold in the slider window, then move the cursor near one or the other of the vertical lines to 'catch' it, and then drag the vertical line to change the contrast.
C. Image statistics
Noise due to counting statistics is apparent in the SEM Magnification Standard image. To get an idea of the extent of the noise, use the MLx - Show Values -> Rect -> Value Stats command.
Slowly drag a rectangle with the cursor (holding the mouse down). Make the rectangle somewhat less than 1 cm on an edge. After the mouse button is let up, the size of the rectangle is frozen, and it will follow the cursor around the image.
When the mouse button is pressed, the average value of the pixel intensities within the rectangle will be shown along with other statistics in the windoid at the bottom right of the screen. It is interesting to note the values for the mean/variance, which should be close to unity for purely Poison statistics and greater than that for image features. In the smoother area (at the top of the image where the indent is, but not including the indent or the dark blemishes scattered around this area), it is around 1.2, in the rougher area surrounding the straight lines, it is about 1.4, and if a part of the indent mark is included within the rectangle, the mean/variance is 10.0 or greater. The thin straight lines do not seem to affect this statistic. Incidentally, the values reported are for the original image and NOT for the scaled or equalied image that may be in the window.
Clicking on the lower right corner of the image window will erase any left over rectangles.
Click the go away box on the windoid at the bottom right to stop the command.