ENHANCE, the man in the cop show says. And the computer geek
"enhances" a picture to several hundred times its original quality.
Pretty silly stuff. But there is a trick or two that haven't been tried.
It should be possible to compensate for optical distortion of pictures
or reflections in pictures
with this algorithm.
Use available lines to try to determine most likely depths of field.
(for example if a refrigerator is clearly seen one can assume its lines
form 90% angles and follow those lines into a virtual background)
This can include looking up the sizes of commercially available items
that may be in the picture.
Having guessed at a field depth, guess at a distortion factor based
on available data (For example: What is the diameter of this coffee
mug)
Using this data guess the most likely shape of unclear objects using
OCR for letters and numbers and object recognition for other things.
If the distorted form of a letter or object can be recognized the
mathematical transform that will bring it into proper shape will also
bring nearby objects into proper shape.