honest question: why is this needed? I would have thought that when downsizing an image, most programs would (should?) use bicubic interpolation to give a pretty accurate downsampled version of the image.
This shows an example of what can happen when you perform a basic resizing of faces. You will notice the distortion when this library is not used. When I had to resize 40,0000 photos, I noticed better visual outcomes when using this library for my testing set of photos.