Dear all,
I am trying to figure out one discrepancy I cannot explain.
If I read in a mask with say direction identity and I linearly interpolate it I get slightly different results than when I interpolate this mask with direction -1 -1 1 (on the diagonals of the matrix). Why is that the case?
To my understanding, the image orientation should not make any difference, right?
Thanks for your help
If you were inverting identity matrix, it shouldn’t. But you probably have slight deviation, and just inverting the diagonal values does not retain ortho-normality of the overall matrix. Meaning, you are slightly skewing the matrix, and as resampling is done in physical space it is affected (with some proportionality to the magnitude of the deviation). Try inverting all the values in the direction matrix and see what effect does that have.