Advice on improving a analysis on images of nanoparticles
Adam Hughes
hughesadam87 at gmail.com
Fri Nov 15 12:02:26 EST 2013
Dear Scikit-Image (SKI) team,
I have over a hundred scanning electron microscope (SEM) images of gold
nanoparticles on glass surfaces, and I've generated several scripts in
ImageJ/Python to batch analyze them. The analysis is fairly crude, and
consists mostly of users manually thresholding to make a binary image,
applying a simple noise filter and performing ImageJ's particle counting
routine. Afterwards, my scripts use Python to do plotting, statistics and
then output .txt, excel and .tex files. Eventually, I'd like to remove
the ImageJ portion altogether and refactor the code to use SKI exclusively;
however, for now, I am mainly interested in improving the results with some
features of SKI.
The images are in a .pdf can be downloaded directly here<http://www.4shared.com/office/xX37rOfh/SCIKIT_TEST_Notable.html>.
(Just a hair too big to attach)
What I'd like to do is to look at a subset of our images and see if SKI can
enhance the image/remove defects. I've chosen 10 images to represent
various cases and attached a summary via googledrive. The images are
categorized as follows (preliminary questions are in blue):
*NICE* - Image is about as good as we can get, and shouldn't have many
artifacts. Can these be further enhanced?
*LOWCONTRAST *- Can the contrast in these images be enhanced automatically
in SKI?
*NONCIRC* - Particles appear non-circular due to stigmation offset in
microscope. Is it possible to reshape them/make them more circular?
*WARPED* - Images that have artifacts, or uneven contrast, due to
aberrations in SEM beam during imaging. * I'm especially interested in
removing uneven contrast.*
*WATERSHED* - These images have overlapping AuNPs, and I had hoped that
SKI's watershedding routines might help disentangle them. The *watershed
segmentation guide*<http://scikit-image.org/docs/dev/auto_examples/plot_watershed.html>indicates that there are several ways to approach this problem.
On the attached PDF, each page shows the original SEM image (converted
from highres.tiff to png), a binary image, our manually chosen adjustment
threshold, and two estimates of the particle diameter distribution (don't
worry about details of this).
I was really hoping that some SKI experts would examine these images and
suggest some algorithms or insights to address the aforementioned concerns.
*The overall goal is to survey the most common problems in SEM imaging of
nanoparticles, give examples of each, and demonstrate how SKI can improve
the particle counting.*
Thanks for you time, and for making a really nice open-source package.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/scikit-image/attachments/20131115/67171a23/attachment.html>
More information about the scikit-image
mailing list