How to measure grain size using ImageJ software

copy & paste from http://osdir.com/ml/java.imagej/2006-04/msg00010.html


---------------------------------------------------------------------------------------------------------------

Anneliese writes,

>
> Date: Sat, 1 Apr 2006 14:46:34 +0200
> From: Klughammer GmbH
> Subject: ImageJ and Metallography
>
> Dear users,
>
> has anybody already made some experience with ImageJ and metallography?
>
> I am looking for
>
> - grain size measurement
> - graphite morphology
> - nodularity measurement
> - particle size distribution
>
> Anneliese
>
And Noel replies.
Refer to the following books.
Computer aided Microscopy by John Russ,
Quantitative Stereology by E. E. Underwood,
and the paper by D.C. Sterio 1984 Journal of Microscopy. (the unbiased
estimation of number and sizes of arbitrary particles using the disector, J
microscopy. 134, 127.

Grain size measurement in metallography proves to be difficult because it is
usually next door to impossible to produce a perfect polish and etch which
reveals all boundaries with sufficient difference to the matrix.
Thus thresholding will be incomplete.
Once you have perfect boundaries, then life is easier.
By perfect boundaries I mean black boundaries on a pure white background.
You may measure grain sizes using all three methods easily.
Linear intercept,
Triple point counting,
And area measurement.

Here are some formulae which are commonly used. (From Russ. J.C. Computer
Assisted Microscopy, Plenum Press 1990, isbn 0-306-43410-5.
P 225.
ASTM grain size G
Using intercept method

G=(-6.6457Log[base10](1/PL))-3.298

Where PL is the number of points per unit length, measured in millimeters.

Or if you measure areas of the grain bodies,

(From Russ. J.C. Computer Assisted Microscopy, Plenum Press 1990, isbn
0-306-43410-5.
P 225.
G=(3.22 Log[base10](NA.M^2))-2.95

Where NA is number of grains per unit area on a polished surface at a
magnification M.

If you count the nodes where triple points are then G is got from
(From Russ. J.C. Computer Assisted Microscopy, Plenum Press 1990, isbn
0-306-43410-5.
P 147.

G=(log[base e]((Nodes/2)-1)/Area)/(log[base e](2)) -2.95.

Some investigators have found it quicker to produce a photo, and to trace
the boundaries out manually using a transparent film and a felt pen. The
resulting drawing is then scanned into the computer and processed in ImageJ.
This will be black and white and easily thresholded into a perfectly
segmented binary image. Such an image is the ideal to which your preparation
must aspire.

Otherwise it may be more time consuming.

If your specimens are suitable, and your etching superb, then
it may be possible to produce a perfect image via the image processing tools
in ImageJ.
Techniques to investigate include, thresholding, subtract background, find
edges, skeletonize, erode, dilate, open, close, watershed and so on.
Also useful may be techniques which differentiate between rough and
smooth, or surfaces with different textures.
Each type of material will have its own behaviour, and you must discover
this for yourself.

I have just talked about grains here.
The particle size distribution will be more straightforward, so long as the
particles are easily discriminated from the matrix. Just be careful that the
smallest particles you need to measure are "larger" than the resolution
limit.
The analyze particles menu in ImageJ will be what you use here.

And for Graphite morphology and nodularity, I have no experience.
But I am sure the methods to be used will have the parameters you need and
these will easily be employed in a macro.

Regards
Noel Goldsmith

Noel Goldsmith
Aircraft Forensic Engineering
Air Vehicles Division
DSTO
506 Lorimer Street
Port Melbourne
Vic 3207
AUSTRALIA
Phone (613) 96267538
FAX (613) 96267089
Email noel.goldsmith@xxxxxxxxxxxxxxxxxxx