So, I finally managed to test it properly on one stack I have. The result is amazing, at least in comparison to Fiji’s StackReg-Plugin. However, it currently takes much more time (Fiji ~ 1.5 minutes, Julia ~ 70 minutes). But this can be due to the @show macro inculded in the sample code.
@tim.holy: Is there any included command to actually save the tfm-values for each slice to a matrix/file? Otherwise, I will just write them to a file individually, so that I can load them later again.