Can an image of unknown text, with resolution of about one or two letters in real life represented by a pixel (like if the image is a photo taken from far away), be used by machine learning or AI to reconstruct the text based on the color of the pixel?

Meaning, if a certain group of letters make a certain colored pixel based on lighting and location, time, distance, etc., can that be reversed engineered to get the original text?

Edit: added if image was a photo taken from far away

  • $\begingroup$ I think it's safe to say no. At least not at this time. One pixel is not enough. $\endgroup$ Commented Dec 23, 2022 at 0:23
  • $\begingroup$ Maybe the exact scene would have to be re-created exactly? $\endgroup$ Commented Dec 23, 2022 at 2:53


You must log in to answer this question.