RescaleIntensity & EnhanceOrSuppressFeatures

Hi all,

I’ve seen lots of posts about the rescale intensity module, but the thing that keeps confusing me is why it is beneficial to do that. Does it improve the actual image quality (which with my limited knowledge would seem weird as just rescaling cannot create new information… or can it?) Is it just easier for the algorithms for some reason to work with stretched intensity histograms?

We’re in the situation that our camera produces 12-bit images which are stored as 16-bit, but even after the 0.0625->1 rescale option, I was surprised to see that the upper limits in the histograms of the rescaled images is very low (~0.15) in most cases. Do the upper limits in the x-axis automatically coincide with the highest observed intensity? I tried to zoom into the histogram pics to see what’s going on there but it didn’t work…

My other question is whether there is some kind of background correction module? We’re using the cellomics system and in the commercial software which comes with it they’ve got two background correction options (rolling ball and a surface fitting method). I saw that the EnhanceOrSuppressFeatures has the rolling ball method as one of the options, so can that module be used for background correction as well?

Many thanks,


I don’t know if rescaling the intensity helps the algorithms, I’ve used a logarithmic modification, but it just helps mainly to identify objects which ain’t clear in the background.
To correct the background I think the module that is normally used is “apply threshold”

Hi imgen,

Sorry that this post slipped through the cracks. Re: image rescaling - It’s worth knowing that in general rescaling intensity is not a great idea, since automatic thresholding methods can accommodate images that use only a small dynamic range. Typically, we (used to) use it to make dark images more visible (e.g., 12-bit to 16-bit conversion) in a consistent manner without distorting the data. But the RescaleIntensity module has fallen out of use in this way; ImageMath takes up the slack for more general arithmetic operations for other purposes.

(As a side note, the automatic thresholding methods already include a logarithmic transformation of the intensities prior to finding the threshold values.)

Re: “background correction” - I believe that you mean illumination correction in this case. For this, I recommend looking at the CorrectIllumation - Calculate/Apply modules and/or the tutorial avilable from our website.