Question about standardising a set of images to allow unbiased thresholding

Hi all!

I was hoping you would be able to offer some guidance on what the best method might be to achieve what I need. This is my fist foray into the world ImageJ and I look forward to building my understanding!

So I have 6 images from a Scanning Electron Microscope at x100 magnification As you can see on the images, the black deposits or plaques are colonies of attached bacteria.

What I would like to do is try to find out the surface area (absolute as well as fraction) occupied by the bacterial colonies. I did a bit of googling and discovered that thresholding would be the best way to achieve this.

Firstly, is that right?

The method I used was to first load up the picture. Then change image type to RGB stack. After this I went to adjust and thresholding and manually changed the min and max parameters so that “visually” all of the bacteria were covered (red).

Secondly, I need to compare the area occupied by bacteria on the different surfaces. However because the pictures all are coloured slightly differently this means setting different thresholds values for each picture, and as I have just been doing this by eye, it is not ideal. So is there a way I can standardise the approach, to keep it consistent between the pictures? Maybe editing the pictures so they are all the same contrast before thresholding, or apply certain filters?

Any help would be greatly appreciated!!

Thanks Amishra123

Hello @amishra123,

That is one of the possible and easiest solutions, yes.[quote=“amishra123, post:1, topic:1021”]
The method I used was to first load up the picture. Then change image type to RGB stack.
What is the original image type? Grayscale? You can (and it is actually easier) to apply thresholds on grayscale images.

One solution would be to normalize the histogram of your images first. You can use for instance open the whole set of images a stack (File > Import > Image Sequence…) and apply the normalization command Process > Enhance Contrast… (with normalize and stack histogram options).

1 Like

Thank you for the quick reply ! I followed your steps, however it doesn’t seem to have worked quite as well, as I had hoped.

One of the 6 pictures is visibly quite alot brighter than the others, following your instructions has lead to that one becoming even brighter aswell.

Well the pictures themselves are .tiff and don’t have any colour. But I presume they are still “seen” as RGB coloured photos no?

You can check this by looking at the subtitle bar of the image window:

Also have a look at the documentation about the image types available in ImageJ.