Example of resampling to a coarser image grid using SimpleITK

SimpleITKResampling

import SimpleITK as sitk
import numpy as np

## Just something I had lying around
inputImage = "ct.nii"

## Load the image. Must be a simpleITK image for this stage to work!
startImage = sitk.ReadImage(inputImage)

## factors to downsize by, this will take a 512x512x118 image to 128x128x118
factors = [4, 4, 1]

## NB, if you wanted to resample to a uniform number of slices, you could figure out what the 1 at the end should be


## Calculate the new pixel spacing and image size
new_spacing = [a*b for a,b in zip(startImage.GetSpacing(), factors)]
new_size = [int(a/b) for a,b in zip(startImage.GetSize(),factors)]

## do some in-plane blurring
sig = 2.0/(2.0*np.sqrt(np.pi)) ## this comes from the Nyquist-Shannon theorem: https://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem
## Blur in each plane separately (NB; there may be a more elegant way to do this, but I didn't try)
blurredImage = sitk.RecursiveGaussian(startImage, sigma=sig*new_spacing[0], direction=0)
blurredImage = sitk.RecursiveGaussian(blurredImage, sigma=sig*new_spacing[1], direction=1)

## Build the resampling filter, and then apply it
resampleFilter = sitk.ResampleImageFilter()
downsampledImage = resampleFilter.Execute(blurredImage,## thing to resample
                                            new_size, ## size to resample to
                                            sitk.Transform(), ## How to transform - in this case, identity since we're just resampling on a coarser grid
                                            sitk.sitkBSpline, ## could also do sitk.sitkLinear which is faster, but worse
                                            startImage.GetOrigin(), ## Keep the same origin
                                            new_spacing, ## The new pixel size
                                            startImage.GetDirection(), ## Keep the same slice direction 
                                            0, ## Default pixel value 
                                            startImage.GetPixelIDValue()) ## output pixel type

## Write the result to check
sitk.WriteImage(downsampledImage, "test_dsCT.nii")


## At this point you could also get the numpy array and go do whatever you like with it.