Supress some undersirable spots

image-calculator
fluorescence
fiji
segmentation
Tags: #<Tag:0x00007fd540365db8> #<Tag:0x00007fd540365c78> #<Tag:0x00007fd540365b38> #<Tag:0x00007fd5403652f0>

#1

Hi All,

I try to find a solution for put off the spot on this image. This is some cell fragment, which burn totaly my picture.

I tried to do a threshold for isolate the undesirable spot and do an image calculator process.

With a difference, I obtain this :

I believe that when the difference is done, the negatives numbers are transformed in positive number…

How can I solve this probleme?

Thanks


#2

Your question relates to fundamental ethical questions in imaging.

It is important to ask why you want to remove the spots.

If it is to enable segmentation or some other kind of analysis, then it may be ok.

If it is to make the image look “better” and you aren’t making a work of art or a journal cover, then you should probably not remove the spots.

If you do succeed in removing them you must document your manipulations.
Otherwise you are presenting data that may not be reproducible, because your image is presenting a view of the sample, the antibodies, etc., that is not representative.

Have a look at these pages for more information.
http://swehsc.pharmacy.arizona.edu/micro/digital-image-ethics

and

If upon consideration you feel it is appropriate to remove the spots, you might look into frequency bandpass filters. The difference function gives you the absolute value of the difference so indeed your values are all going to be positive.

Hope this helps.


#3

Thanks for your recommandations,

I am testing a new protocol, that will be modify for supress the spot before the imaging.
In parallel, I try to find antibody which could be used for the futur experiment. This is why I want to supress the spot, just for see the antibody fluorescence and the localisation. Thoses spots are in all my channels.
The aim is also to see if my protocol don’t hurt the structure… but with thoses spots, the 3D reconstructions are very hard!

I will try the frequency bandpass, thank you!