Dear Torsten,

Many thanks for your explanation.  It is very helpful.

Let me try to explain my problem more clearly.  I've a reference image name, lobus2.nii and a floating image name lobus1.nii.  I've successfully applied the WARP to get a warp xform.  Now I would like to do the scalar measure to export the X-,Y-,Z-component of the warp xform.

It worked fine when I use the below example, the component will be x-component, y-component, z-component.
xform2scalar --mode component lobus2.nii warpxform.list

The above results are all correct.  The deformation filed (combined X-,Y-,Z-component) could deform the lobus2.nii(reference) to the lobus1.nii(floating).  I would also like to get the deformation field from lobus1.nii(floating) to lobus2.nii(reference).  So I use the below.
the wrong formula I used at begining: xform2scalar --mode component lobus2.nii -- --inverse warpxform.list
The fixed formula you suggested: xform2scalar --mode component lobus2.nii --inverse warpxform.list

The fixed formula could export not all-zero result.  However there are lots of Inf.  I guess it is out of domain.  However in my problem, the initial bounding box of lobus1.nii and lobus2.nii is the same.  And the transformation is identity.  

The two model is almost the same.  I do not think it is the problem. of type a).  I will try to trace the program to rule out the problem. 

Many thanks.

Jaguar