Is it possible to prepare sample without such aggregates? By my experience, due to a high number of aggregates you should do it manually, by the Tools-Distances. Authors of Gwyddion do not recommend it (see Article Study of user influence in routine SPM data processing
), due to a low reproducibility of results, but here usual ways of grains masking (Grains-Mark by edge detection , for example) seems not to be effective.For the start, try to measure manually for every particle the lengh once, and the diameter N times. Do not forget to remove background before the measurements.
to Ebrahim Jalali Dil:
Would ImageJ be helpful in automatic aggregates recognition and parsing?
I suspect that any image analysis package (including Image J) would have issues with such a picture - contrasting background (so no unique thresholding value) and overlapping particles just to start.... Only a manual deconvolution (picture + ruler) would be feasible. I'm still needing an answer to my originally posed question. What characteristic dimension is required to be measured for particle size? Maximum/minimum length? Circular equivalent (disc with same number of pixels as the indicated particles), volume equivalent (number to volume conversion). Is the scale on your axis '5 microns' for the (full) x-scale. Is the y-axis also the same? That is, has the system been calibrated in both x and y directions. Important if you're viewing on a 16:9 or 4:3 monitor where they'd appear very different...
Thank you so much for your time and reply. I need to calculate the maximum/minmun lenght and width of the nanocrystals. As it was a JPGE image so scale is not proper here. Please see the attached original AFM image. Thank you once again!
The AFM 'image' is in Notepad and just a bunch of letters and numbers. You need to post as a jpeg (or preferably a tiff). I think there's a sufficiently small number of particles that you can do this by hand and with a ruler. I don't believe there's an automated route in your case, for reasons I outlined previously, but I always learn more by being wrong...
Thank you for the raw data.
Just for example, I tried to mask the particles - by Otsu's algorithm, for example ("Data Process -- Grains - Mark by Otsu's"), but maybe "Mask-Mark with..." would works better.
I calculated the distribution of minor semiaxes of equivalent ellipses (it should corresponds to the particles width, meanwhile the major semiaxes - to the length) "Grains-Distributions..." and it seems fine - see test2.png image, but actually there is the problem, which I mentioned before. Some of snow-like aggregates have been counted as a single particles and you can see some false big particles, especially on the higher resolution of histogram (test3.png) .
However, results seems much better that I've expected. I suppose, you should try to figure out the best masking algorithm and after that to plot a necessary grains distributions.
to Alan F Rawle:
It is a raw AFM data with the true scales and it looks good in the mentioned Gwyddion software. But what do you think about applicability of the equivalent ellipses model in this case?
Vasily Lebedev :
Thank you so much for doing anlysis of AFM image. Yes, I have also seen it. Some of the particles are not well distributed. Once again thank you for the time and suggestions.
Nice work, Vasily. Equivalents will work in every case - this is a double-edged sword though. The poster was looking for maximum and minimum length and with the small number of particles, I felt that this was feasible by hand. It may be that converting the ellipses into equivalent pixel areas (the area of an ellipse is easy to calculate) would allow circular equivalents to be obtained.
It may be worth looking at the sensitivity of the generated result bit simply masking a few 'particles' and evaluating the outcome. In imaging it's easier if particles or regions can be well-separated for obvious computational advantages.
I have other routes available to me with a tiff (or a converted to 256 gray scale gif) to compare - obviously I don't work with the Gwyddion software but I'm very familiar with Image J.
I prepared the tiff image after some usual AFM tricks, such as background removal etc. Size of the image is 5x5mkm.
Also I tried to mark particles using its height (brightness), and to remove small (few pixels) false-positive particles from the distribution using threshold values. Final distributions (50 points histogram) of major and minor semiaxes are attached.
Unfortunately, for now I have no idea about the reproducibility of the obtained data. Tuning of the threshold level does not affect on the final distributions, but I do not know, what happens in the case of the use of another masking algorithm.
Dear Vasily ,
Thank you so much for the help.
Wish you have a good time:-)
- Badges
- Science topic