I am searching for a good character extraction method, or sometimes it is called stroke-model or stroke filter. So, I;ve seen many papers, but they all take a long time for understanding and implementation, I want to ask if someone knows some good source codes or demos? Also I want to get some kind of full overview of methods available on these theme : character extraction from images, (grayscale). The main problem is to get a regions of image that include only characters and then some binarization can be made. After that the feature extraction is done (actually OCR works then).
A:
Maybe GNU Ocrad can be interesting? I haven't looked at the source though.
Emil Vikström
2010-05-14 08:06:50
interesting and open source project, I try to analyze the source, thanks!
erjik
2010-05-14 09:59:43
A:
An area with characters is recognized by a large number of sharp edges. There will be some preferential directions, but this is not as strong as you'd see with box shapes.
You seem to assume that it is possible to get "regions of image that include only characters". This is too optimistic. Just look at this very page. There are symbols mixed in with text. And above this editing box, the first four toolbuttons are B, I, a globe and ". Five, if you count the thin divider bar | after the I
MSalters
2010-05-14 08:15:17