Re: HELP!!!-segmentation and quantification of DIC images

Posted by Jim Peterson on Feb 28, 2020; 1:14pm
URL: http://imagej.273.s1.nabble.com/HELP-segmentation-and-quantification-of-DIC-images-tp5022977p5022984.html

I was a little sloppy with the Gaussian blur.  You may be able to reduce the
radius or remove it all together, and may get more accurate resultsget
better results.

As for removing the shadows, I would separately identify the shadows and
then subtract them from the results from the original segmentation.  A
simple thresholding using Otsu with light background seems to do a decent
job of finding the shadows.  In the upper-left and lower-right corners you
pick up a little noise in the background due uneven illumination, but it
shouldn't be a problem since it is in the background.

<http://imagej.1557.x6.nabble.com/file/t381520/Screen_Shot_2020-02-28_at_7.png>

So I would suggest then the following:
    1. Do the segmentation described previously, resulting in a binary mask
of the objects plus shadows.
    2. Segment the shadows using simple Otsu threshold, resulting in a
shadow mask.
    3. Subtract the shadow mask from the original mask.
    4. Run particle analysis on the result to generate object properties.

- Jim



--
Sent from: http://imagej.1557.x6.nabble.com/

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html