CorrectIlluminationCalculate - what do the values in the .mat file represent?



I created an illumination function using the module in the title above - it is here: IlluminationCorrection.cpproj (642.0 KB)

Illumination function - Background.
Rescale - No.

The output illumination function is a ‘.mat’ image. What do the various values for each pixels represent? Are they the absolute background intensities of my images? Or is it a percent/ fractional value?

For eg. most of the pixels of the illumination functions image are in the range 0.03 - 0.05. Is this value subtracted from my ‘sample’ images? Or is it 3-5% of the maximum or such?


There’s actually an error in the module help (which you just helped us realize, so thanks!), so it’s no wonder you’re confused.

With Rescale turned off, it’s an absolute value representing the value of the least dim pixel in that position across all of your images (that you should use with ‘Subtract’ in CorrectIlluminationApply); with Rescale turned on, it’s a number >1 that’s the value at that pixel relative to your dimmest pixels in the image (for which you should use ‘Divide’ in CorrectIlluminationApply). Does that help at all?


Thanks for the quick reply.
Indeed, that helps, but I have a followup question.

When I open my ‘sample’ images (they are .tif files) in matlab, my ‘positive’ intensities (i.e. intensities when the pixel belongs to a nucleus - which have been stained, as opposed to a background) are approximately 60,000.

Does this mean my microscope is illuminating uniformly, because the background values seem to be very small compared to positive values?

Or does cell profiler change the image intensity values to a different scale of its own?


CellProfiler rescales everything 0-1 based on the bit depth of your image, so the 0.03 and 0.04 represent 0.03 and 0.04 of the bit depth.


Thanks for the clarification.

By bit depth, you mean the range of intensities in my image set? max-min?
I looked up image bit depth, but that seems to be something slightly different.


Bit depth means the maximum value that COULD be in your image set based on how your camera stores data- it’s usually 255 (8 bit), 4095 (12 bit), or 65535 (16 bit).

If your average nuclear intensity is 60,000, then by process of elimination you must have 16 bit images (you’re also running close to saturation, which if you’re hoping to do quantification is a bit risky); your average background pixel according to CP is 65535*0.04=~2600.