Quality of Screen Images Recognition and Text Symbols

[hypos...@gmail.com]

Hi,

Are there any tips to improve quality of text recognition
from screen-captured images besides re-scaling them to 300 DPI?

Any preference to image formats for a capture that could be
helpful for tesseract?

Yet another question is how to handle i.e. "step-by-step" with two hyphens
that are recognized as letter 'r' somehow?

Well, a similar case is some text with "Menu->Edit"
and how to handle underlined web-links i.e. from PDF docs?

Thank you in advance for sharing of your ideas!

Hypo