zoom_image with nearest neighbor "scaling"

[AnderBiguri]

For a particular thing, I needed to use zoom_image to do nearest neighbor scaling (this is a bit of a misnomer, perhaps "None" scaling is a better name?). Essentially I am doing zoom_image on an ImageData that contains labels, and thus they need to keep their pixel values unchanged.

I have modified zoom_image such that allows this. I post it here in case this is not a "just for me" solution, and there is interest on having it in some form or another in SIRF. If the answer is "no", thats great, we can close the issue but have it as reference. If its yes, let me know, I can do a PR.

My zoom_image reads:

def zoom_image(self, zooms=(1., 1., 1.), offsets_in_mm=(0., 0., 0.), size=(-1, -1, -1), scaling='preserve_sum'):
        """
        Returns a zoomed image.

        All coordinates and indices are given as (z,y,x).
        To leave the size unchanged in any dimension, set the
        corresponding size to -1
        Supported scaling options are: 'preserve_sum', 'preserve_values',
        'preserve_projections'  and 'nearest_neighbour'
        """
    zoomed_im=self.clone()
    if not isinstance(zooms, tuple):
        raise ValueError('zoom_image: zooms should be tuple')
    if not isinstance(offsets_in_mm, tuple):
        raise ValueError('zoom_image: offsets_in_mm should be tuple')
    if not isinstance(size, tuple):
        raise ValueError('zoom_image: size should be tuple')
    if not scaling in ['preserve_sum', 'preserve_values', 'preserve_projections','nearest_neighbour']:
        raise ValueError('zoom_image: scaling should be preserve_sum, preserve_values, preserve_projections or nearest_neighbour')
    
    np_zooms = numpy.asarray(zooms, dtype=numpy.float32)
    np_offsets_in_mm = numpy.asarray(offsets_in_mm, dtype=numpy.float32)
    np_size = numpy.asarray(size, dtype=numpy.int32)

    if scaling=="nearest_neighbour":
        # STIR does not do this. so we need to make a template with the size we want, then call the resampler. 
        try_calling(pystir.cSTIR_ImageData_zoom_image(
                zoomed_im.handle, np_zooms.ctypes.data,
                np_offsets_in_mm.ctypes.data, np_size.ctypes.data, 'preserve_values'))

        resampler = reg.NiftyResample()                         # create resampler
        resampler.set_interpolation_type_to_nearest_neighbour() # nearest neighbour
        resampler.set_reference_image(zoomed_im)                # Template of target image (data will be ignored)
        resampler.set_floating_image(self)                      # input image
        resampler.process()                                     # do resampling 
        zoomed_im=resampler.get_output()                        # get output. 

    else:
        try_calling(pystir.cSTIR_ImageData_zoom_image(
                zoomed_im.handle, np_zooms.ctypes.data,
                np_offsets_in_mm.ctypes.data, np_size.ctypes.data, scaling))

    return zoomed_im