Registration/alignment of two images

[markbuckup (Mark)]

Hello,

I am trying to find the optimal registration process for these two images. They are of the same tissue across different stains, which means that they have different RGB qualities but are of the same basic size. There is some tissue damage in the second image which is causing issues for me when I try ITK translation registration. My initial goal was to do a composite registration consisting of a translation or affine shift, followed by a Bspline transform to account for the micro-tissue movement. Any idea on if this is the best approach? I haven’t had much success.

mark

[dzenanz (Dženan Zukić)]

Mutual information metric should be able to account for color differences. But your images have such different features that it is no wonder registration struggles/fails.

You could incorporate manual intervention for problematic pairs, as suggested by @grothausmann.roman in this paper.

[matt.mccormick]

It may be helpful to pre-process the images first with a filter like the ones provided by ITK Structure Preserving Color Normalization.