Although the created geometry is a non-image geometry, it's origin is left in the (0, 0, 0) mm position, and the bounding box lower bounds are negative to compensate this. Moreover, the spacings are not permuted according to order of the input geometry axes, therefore the upper bounds are also incorrect if the input geometry is an image with anisotropic voxels.
Description
Description
Event Timeline
Comment Actions
As soon as we have an image spacing other than 1, we have indeed an error in our calculations. Everything else seems to be the difference between object space and global space, which is why the missing offset was "okay", though.
The calculation in your fix is correct.
We still have to check the permuted thingy, though.