[exwgfgab] Visual data compression

You are given a pixel raster of a specified size onto which you can write any RGB image.  A human trained to read the image will then read off data encoded into the image.  What is the maximum amount of information one can fit into the image that an unaided human can read?

Assume (or not) that the human can magnify the image to see the individual pixels.  Probably depends on the color sensitivity of the eye which is tricky because it is affected by adjacent colors.

Another formulation permits vector graphics of a given physical area, and the human is not allowed to magnify.

Instead of abstractly information, what if we want to encode text in a given language?  We want to do better than simply rendering the text as an image, yet still retain the ability for a human to read it unaided.