ImageJ sometimes outputs a 32-bit TIFF (16-bit range of values but floating-point instead of integers), such as through the Z Project>Median function. When this image gets loaded into CellProfiler, it does not rescale it as it does with <=16-bit images. While this is easy enough to correct by diving the image by 65535, it does introduce an inconsistency that can take a while to track down.
I attempted to upload the image, but Firefox froze up on me. You can make one yourself by making a Z-stack in ImageJ and then Image>Stacks>Z-Project…, selecting “Median”, then save as TIFF.
On that note, I think it would be helpful to be able to specify, in the LoadImages (and LoadSingleImage) module the number of bits the data is encoded in when “Rescale Intensity” is unchecked, rather than just rescaling to the file type’s bit depth. All of my images are collected as 12-bit-in-16-bit, and they are treated as 16-bit by the image loading modules. On that note, having the check box be labeled “rescale intensities?” is a bit misleading, since it is really toggling between two modes, rather than rescaling and not rescaling. I’ll look into making these changes myself, but do let me know if I’d be off-base in doing this .