I need to remove artifacts (dust or other debris) from other images taken by the same microscope.

I am collecting images from our microscope to attempt to train a cnn to recognize how out of focus the microscope is (this is the article if you are interested, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5943362/). However, there are artifacts on our images as you can see below because of imperfections on the camera and possibly internal microscope lenses. Below is first, a picture taken without anything on microscope and second is a picture of the cells we are working with. Is there any method I could try with matlab to attempt to remove these background artifacts? We are having trouble training the network with our current data (we have done it before with another microscope/camera set that worked quite well) and we think this is the problem. Thanks in advance for any advice!

Without Cells:

With Cells:

Answer 1

[Raghunandan V]

Hi,

This is an interesting question. Can you try using edge detection filters and then send it through CNN?.

[Brae Petersen]

Thanks for the response! Do you mean using edge detection on the image without cells to create a mask of sorts where the debris are?

[Raghunandan V]

Yes, I see that debris edges are much thicker than that of cells. If you can set the threshold to find the debris then you can eliminate it by passing through a low pass filter.