Image analysis of barley culm sections

Hello everybody;
This is my first post on this forum, and I’m not really an expert of image editing/analysis.
So In order to analyse many sections of barley straw we decided to go for Fiji/ImageJ software. The samples (which can be considered small elliptical shapes) are attached to a black cardboard and then scanned. After this I found-out a way to reduce the RGB image to a binary image without loosing data–> in this way the “magic wand” tool as “analyze-particles” should be able to identify the shape of the sections. But unfortunately analyze-articles doesn’t work really: while few samples are recognized by the software others are completely ignored even if they match the AREA and CIRCULARITY parameters. Is it a sort of bug or what?

Hello @bretha and welcome to the ImageJ forum!

It seems you are having trouble setting the proper parameters or maybe the binary image should be inverted (do you have black or white samples?).

It would really help if you post here an example image and the results you are getting.

Cheers!

1 Like

Hi @iarganda


here an examle sheet of my samples. So yes I have black samples, actually I didn’t try to invert them to white but imho I don’t think it will change the results (but maybe I’m wrong, I’m really new to this kind of analysis).
Sometimes the software tells me that threshold parameters are not correct, but honestly I do not know what it means (I always thought that threshold could be a problem in complex RGB images not with this kind of binary images)

As you can see few of them don’t have a continuous perimeter but as I told you is just an example. Usually all the sections look nice.

I see! Analyze Particles works on solid objects with distinct background. For example:

In your case, you have the borders in black and all objects in white (including some we can consider background). An easy way to proceed would be to first label all white objects using for example the Connected Component Labeling from MorphoLibJ (Plugins > MorphoLibJ > Binary Images > Connected Component Labeling). This is how the result would look with the Inverted Glasbey LUT:

Then you can simply take morphological measures of the different labels using MorphoLibJ’s Region Morphometry plugin (Plugins > MorphoLibJ > Analyze > Region Morphometry):

Then you just need to filter the table removing the values of low circularity and very small or large area.

I hope it helps!

3 Likes

@iarganda thank you very much for your nice suggestions :slight_smile: it seems working

1 Like