This approach sounds reasonable to me (I may be biased as I pointed you in this direction).
You are doing well for someone who has just started with medical imaging, hopefully the following will move you further along:
The main difference between medical images and other types of images is that they are
spatial objects. That is, the CT image occupies a specific region in space with a
known size (think of it as a box in 3D space).
This brings us to your question with respect to grey slices. The reference_image represents
a box of certain dimensions (sizes*spacings), the largest size of expected patients, so that everyone will
fit into this box without distorting their physical size. When you have a smaller patient they fit in the box with space to spare (grey slices). As a sanity check you can measure structures in the original CT and in your resampled CT and see that their size remained the same (fewer/more pixels but larger/smaller spacing between them).
As you are working with CT: during the initial code development it is good that you get these
grey slices, showing you where the original data existed and where you were looking outside
of the original (grey). When you are happy with the code you should switch to a more appropriate
value, in the case of CT, a value of -1000 (Hounsfield value for air).
Looking at the code, you have:
resampler.SetDefaultPixelValue(img.GetPixelIDValue())
This is incorrect usage of the SetDefaultPixelValue method. The GetPixelIDValue referes to the type of the
pixel which unfortunately is a number and thus accepted by the method without complaints. If you don’t set
it, the default is zero, any arbitrary value you use is fine (including whatever GetPixelIDValue returned).