Re: Standard set of image pre-processing filters for poorly hand printed images?

On Nov 8, 12:07 pm, aruzinsky <aruzin...@xxxxxxxxxxxxxxxxxxxx> wrote:
There are filters to fill in broken lines in fingerprints which you
should Google.

On Nov 5, 3:38 pm, Milind Joshi <milind.a.jo...@xxxxxxxxx> wrote:

Hi,

Does anyone know of a standard filter or filters applied in sequence
to a B/W hand printed images to fill in broken lines, etc., without
affecting those characters that are well-written with complete
strokes?

We have a bunch of images with different writing styles, applying a
single set of filters on all of them deteriorates quality on those
images that were good.

Also, applying filters and checking results to see if the filter was
OK is too time-consuming.

Standard thickening/smoothing operations built into OCR/ICR engines
didn't work very well, so we're going to have to write a custom image
pre-processor for these images.

You can find a sample athttp://www.ideatechnosoft.com/sci.image.processing/2007_11_05_sci.ima...

There are many more samples, so that should not be a problem if you
want us to try out suggestions on a wider set of images.

What would be a general approach to arrive at a good sequence of image
processing filters or to choose the best one for the job? Is there a
filter out there that is recommended for this type of problem?

Any help, pointers, reading material, manuals, expert suggestions
would be greatly appreciated!

Best Regards,
Milind

That is actually quite a cool idea. We never looked into fingerprint
recognition image pre-processing algorithms before, though there is a
kind of regularity in fingerprint lines that is not found in
handwriting... the general idea of extrapolating lines to "fill in the
gaps" is a good one.

Regards,
Milind

.

Relevant Pages