ralston@cybavax.UUCP (R. Alston Elec. Eng. Guest) (08/13/85)
A friend of mine is working on a small character scanning device based
around a CCD sensor for his blind girlfriend. The system as it stands
will return a 24 (vertical) by 6 (horizontal) bit image of what
appears under the scanning head. The head is moved across a printed
page by hand , relying in the operator to keep it moving parallel to the
lines of text. What he would like is some ideas on how best to process
these bit images to find out which character is being scanned.
Some assumptions that are made:
1) The system will always be used on printed books or papers
and if it has to go through some kind of learning stage for
each given typeface that is acceptable.
2) There is variable magnification on the scanner so that it
can be adjusted so that the largest character (capital M or W?)
will be contained within the 24x6 bit array.
Now I can forsee problems just in the fact that the scanning system is
moved by hand across the page eg. how do you decide that a complete
character is contained within the array? The system will be connected
to a small micro (a BBC in the first instance) and apparently the
CCD stuff returns a complete scan every 4.5 ms. Any ideas, suggestions
or pointers to algorithms etc. would be gratefully recieved and I will
pass them on.
Thanx in advance for any help
Robin Alston
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
login name : ralston machine name: cybavax
uucp address: {UK}!ukc!reading!cybavax!ralston
JANET address: ralston@swxa/234207920018
University College, Swansea SA2 8PP, UK.
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -