As far back as the early 1970's I was first exploring the possibilities of modelling the overall performance of the human visual system - peripheral as well as foveal - and its inference on both the modelling of visual search and the prediction of search strategies. The key to all this, as my colleagues & I saw it, was the finding that the distribution of retinal receptors across the human retina could be well fit by a simple formula
n(2) = n(0)/SQRT(2 + 1)
where n(2) is the number of retinal receptors / degree at angle 2 from the fovea, n(0) is the number of retinal receptors /degree at the fovea and 2 is measured in degrees
The consequences of the foregoing, together with substantial practical experimental support, were presented in Chapters 7.5 & 8.2 of my first book 'Vision and Acquisition', while the implications on development of practical computer vision simulations (as things stood in the early 1990's) were discussed a little in Chapter 18.2 of my second book 'Computer Vision ...'. Sadly even at the time of my second book (as inferred by the said discussion) neither the capacities of computers of the day nor the digital imaging capabilities of the day were remotely compatible with any practical exploration of the possibilities. Nevertheless, at around that time I did develop some basic software which was theoretically capable of carrying out the necessary progressive compression.