Good article, and a much needed one.
If we can come up with a light-weight way of evaluating the frequency of signals in an image, that can be used to switch between vImage resizing and another technique better suited for images with high frequency signals.
Best of both worlds.
vImage may have the best average performance and result. But, what’s even better is choosing the right technique for each image to best suit it. Or, the right resizing technique for each part of an image, if the image is extremely dynamic.