Dear Alexander, Thanks for your reply. Actually, I am doing the reverse process. I am testing some registration methods and I need some nonrigidly transformed images to apply to my registration pipeline. I want only to measure the degree of registration improvement by some simple metrics such as mutual information. First, I used brainweb with RF nonlinearity and noise, but, RF nonlinearity is a kind of intensity nonlinearity and I think is not suitable to test a registration algorithm (Am I right?). So, I need to have a sample image and nonrigidly transformed version (spatially nonuniformed) of it to test my algorithm. Best, Siamak