I am trying to recognise text from a screenshot. I have the benefit of
knowing exactly where a line of text starts, however sometimes
neighbouring glyphs either overlap or even join together with no
spaces between them at all.
I.e. I cant use a simple method of checking that at least 1 horizontal
pixel separates glyph 1 from glyph 2. Some examples include:
"ev" or "ve" : The the e doesn't join with the v but they overlap i.e.
some pixels share the same x coordinate.
"ff" : The two glyphs are "connected" to each other/joined
There are a few other cases also where a pixel is shared by both glyph
1 and glyph 2.
I have been looking with google, but am new to this area and may be
searching for the wrong terms. Can anyone point me to information
about different algorithms i can use to segment the glyphs in a line
of text and especially in situations like the above where there is
sometimes little or no separation between glyphs.
Note: This is for a binary image (i.e. pixel is on or off), though
also i will want to apply the same technique to some other anti-
aliased text that is in the image too.