June 10, 2004
This is a convolution operation, sometimes called 'unsharp masking'.
It amounts to subtracting a blurred (low pass or mean-filtered) image from
the original, and then enhancing the contrast back to full range.
||Convolution kernel for a 3x3 mean filter smoothing. Each pixel is replaced
with the sum (or average) of itself and its 8 neighbors.
||Convolution which does nothing but reproduce the original (at 9x original
||Convolution kernel for the 3x3 sharpen. The first kernel on the left is
subtracted from the second.
|Kernel file in Sample Images / Kernels for getting equivalent
results with Process / Convolve Process /
Filters / Convolve
|| Mean( 3x3).txt
- Open Sample Images (in your lab-images directory) / FESEM / Icextl.tiff
- (Optional, Mac only: Duplicate the image. Make sure these two images are
the only images open. Sharpen one copy, leave the other alone for comparisons.
If the option key is held down while one of the images is zoomed or
the other image will duplicate the zoom and scroll after you let up on the
mouse button. Wayne Rasband, in the NIH Image manual, recommends tiling the
images before zooming and scrolling. This ensures that the windows are the
same size. If this is not the case, the aspect ratios in some of the windows
may change. This will not affect the image pixel values. This dual zoom and
pan enables detailed comparison of the sharpened image with the original.
- Select the Process-> Sharpen menu.
The crystals appear much 'crisper' than in the original. The boundaries
may appear sharper, but are actually not - the original version should be
used for processing, not this sharpened version.
- Edit -> Undo Filter menu. (original restored)
- Select the Process-> Sharpen menu holding down the Option
key. (Not applicable to ImageJ). A kernel
with proportionately less of the original image is used for more sharpening,
when the option key is held down:
- Use Process /
Filters / Convolve / Open and select Sample Images / Kernels
/ SHRP3X3-9.TXT for more "sharpness",
or edit the default filter like
to get the result below.
This degree of sharpening is often too severe.
- Other Sample Images to try.
- Sample Images / Filter Samples / face.tiff See results in the
- The sharpened version might be better for publication. (If enhancement
like this is used for scientific publication, the figure caption or text
should say so.)
- Sample Images / Filter Samples / Viking mars rocks.tiff
- No improvement is seen in the very blurry objects in the foreground
- this is a 3x3 filter which takes advantage of only short-range fuzziness.
- Sample Images / fesem / leafnorm
- This image has fine grain noise, which is what is enhanced with this
- Sample Images / mas journal article / amosite1.tiff