For a side project at home, I have several pdf documents in "standard" form (White background, black font for text). For several pages, I have footnotes which are different in size from the main body. Do you know a way to identify the text block in a document based on the font size?
For most fonts, just digitally smear the image horizontally, then look at the height of the resulting gray and white bars. Assuming typical text the frequency of capitals, descenders and ascenders should be small enough to just contribute a bit of gray.
A more sophisticated method would try to pick out the lines, then find the tips of the ascenders vs. the baseline. That would be the height of the font in pixels (or whatevers). Then find the height of each pixel (or whatever) in points, and you have the font height.