Question regarding masked registration with different shapes


I’m wondering how the sampling works when performing registration with masks.
In the elastix manual its written its sufficient to provide a image for the fixed image when there is no severe noise in the moving image ROI area.

What I’m confused about, is that the mask for the fixed image that i provide contains less channels/slices than my fixed image, but it still works.
For example when having 2 3D volumes and a mask matching the dimensions of the moving image:
fixed image: 100x10x10
moving image: 50x10x10
mask: 50x10x10

So how are samples drawn when the masks for the fixed image has less channels (or more) than the fixed image itself.

I’d be very thankful for any insights!

Hello @kenrickmoke,

This has to do with a fundamental concept of ITK, images occupy a physical/continuous region in space. For more details, see discussion in this post.

1 Like