Scaling up an image transformation matrix for high resolution image registration?

I have two very large, high-resolution images I would like to register. I have more or less followed this tutorial on how to register the images, but in order to do so, I had to significantly decrease the image size.

Here is the pseudo code for what I did:

Create mask of object 1
Create mask of object 2
Down-sample both masks by a factor of 5
Obtain a transformation matrix to register object 1 to object 2

Ultimately, I need to use the high-resolution image for other analysis. Is there a way I can up-scale the transformation matrix by a factor of 5 such that I can get a pretty good registration of the two high-res images, and use that as an initial condition for a final registration step? I tried just using the resulting matrix, and it did not give productive results, as well as multiplying it by a factor of 5, and that did not work either.

[Laxmi Gupta]

Hi, I am using Image Alignment Toolbox for the same purpose. The function 'iat_warp_updown' in that toolbox does exactly what you want to do. Hope this helps! Regards,