Registration between channels



I’m just wondering if there’s a way to retroactively correct for registration between images taken in the red and green channels in CP. I accidentally used a dated skew plot to image, and now my images don’t line up - this is will be a systematic error, so I’m wondering if there’s a way I can add a correction for this (crop the images in different places, and overlay the crops, perhaps?) into my segmentation pipeline. If so, how would I go about doing this?

Looking forward to hearing from you,



Hi Erin,

You can use the Align module to register misaligned images. Let us know if this helps!



Thanks David,

I used the align module as suggested with much success… Most of the time. I’m having an issue wherein when I use the Align module in my pipeline (attached), it works well for images containing cells, but when the image is of an empty (or almost empty) well, the pipeline identifies hundreds of ghost objects along the left hand panel of the image (also attached). Any idea what I could change in my pipeline to fix this problem?

Thanks a heap!

Rad52_1IPA0ISA_ALIGN_WithImages_5811_Pipe.mat (2.52 KB)


Hi Erin,

It seems like the region on the left is bright enough so that IdentifyPrimAutomatic module is able to detect spurious objects. Is increasing the lower threshold bound until these objects disappear an option?



Hi Mark,

I’ve tried that, but it doesn’t seem to work. Also, I’ve realized that in images in which there are very few cells (less than 10), the align module causes the same pileup of ghost objects along the left side, but the identify primary module ignores any actual cells that are present in the image. I’ve created a system for filtering out ghost objects that have an intensity value of 0 (which they all do) retroactively, but the fact that it’s ignoring some actual objects in favour of identifying a bar of ghost objects has become a real issue. I’m attaching a sample image of this circumstance also.

Thanks for your help, I’m out of ideas!



Sorry, let me be a little more clear about that;

The presence of the ghost objects seems to negate the identification of most actual cells in the image - I can increase the lower threshold and remove the majority of the ghost objects, but in doing so, I also lose the correct identification of any cells that were identified originally - in other words, I go from correctly identifying maybe 2 of 10 objects, + 1000 ghost objects, to identifying 0 actual objects, and ~100 or so ghost objects. Not ideal!

Thanks again for your help,



Could you upload a set of images which produce this problem in CP?



We ended up writing a script to weed out the ghost objects - thanks for your help!