Novelty in images

Almost every image that we see contains some content with which we are familiar

Consider an image that might be entirely Novel:

One comprised of random pixels, in which every pixel is independent of its neighbours

This is an unusual image. It would be incredibly difficult to accurately remember this image of around 1000 pixels.

More typically the observer might expect areas of similar texture, separated by extended boundaries, even in the simplest image
(our mind even seeks familiar patterns within the random image above)

Most images that we perceive have some content that is familiar.
It is very difficult to estimate the novel information content of such an image, without knowing the previous experience of the observer

One can however use an image which the observer is familiar, but with a few specific changes, and measure the ability to monitor these changes. A possible test then is to switch between almost identical images and measure the speed at which we can resolve differences