The image origin and orientation tags are read from the DICOM file in cmtk::VolumeFromFile::ReadDICOM, but they are not converted into an image-to-physical space transformation matrix.

Because there are only two orientations currently read (for the x and y axes), we either have to find the missing orientation in the file still, or compute it from the other two directions via vector cross product.

Not a high priority, because we haven't seen a multi-slice DICOM file for a long time. Stacking of single-slice files is unaffected by this.