scott <sco### [at] scottcom> wrote:
> Especially as in the example mentioned, car number plates, you know the 
> font and size characters that are used.

  That's certainly true. The algorithm doesn't even have to guess what
the original shape was, it only has to compare the image (after correcting
for perspective) against the known font and estimate the closest match.
It probably helps if the image is first filtered to remove as much
extraneous noise as possible, and any extra information from additional
frames added.

-- 
                                                          - Warp

Post a reply to this message