Character Recognition

Character Recognition

Post by Ryan Han » Wed, 02 Jul 2003 10:56:39



Hello everyone.  
I'm looking for a little guidance on a programming task that I'm
trying to accomplish.  Part of a program I'm writing has the
responsibility of scanning through blocks of music videos (captured
from mtv2) and determining where in the file each seperate video
begins and ends while extracting the textual descriptions of the
videos that are flashed in the lower left corners of some of the
frames.  I plan on doing some image processing on the frames that
contain the textual information and sending this processed image to
OCR software.  How complex/difficult would a task like this be for an
almost-graduated computer scientist?  Is this approach a feasible one
or should I consider something else?

What I need help with is making sure the image that I will be sending
to the OCR software has characters that are recongnizeable by the OCR
software.  How would I go about researching the image processing I
will need to do in order to maximize the characters readability?

I have very little experience with OCR.  How would different OCR
applications compare for a task like this?  Are some going to be able
to read the characters with less complex processing?  Is it worth
investing in commercial software?  I plan on using GOCR/JOCR (
http://jocr.sourceforge.net ).  Is there reason's I would want to use
something else?

Here are some sample bitmaps to look at
http://www.public.iastate.edu/~hanksr/fbs1.bmp
http://www.public.iastate.edu/~hanksr/fbs2.bmp
http://www.public.iastate.edu/~hanksr/moby.bmp

Thanks in advance. :)
Ryan

 
 
 

Character Recognition

Post by Hans-Bernhard Broeke » Wed, 02 Jul 2003 18:20:48



> What I need help with is making sure the image that I will be sending
> to the OCR software has characters that are recongnizeable by the OCR
> software.  How would I go about researching the image processing I
> will need to do in order to maximize the characters readability?

The simplest, yet most effective image pre-processing trick almost
certainly is to search for largeish areas that hardly change between
frames at all --- music video are full of motion, text insets are
almost perfectly static.

Find such parts, and isolate them by looking for the transition
between high change rates and virtually none.

The OCR itself is a separate sub-task --- you may want to ask about
that in sci.image.processing.
--

Even if all the snow were burnt, ashes would remain.

 
 
 

1. OCR - optical character recognition

Hi folx!

Any people here who can help me finding any information about
programming
techniques to optical character recognition???

Greets
        Ralph

P.S.: reply any responses to my email, I am not reading news often
--
======================================================================

 Eibseestr. 18c          WWW: http://www.fh-augsburg.de/~raresoft/
 86163 Augsburg          IRC: rr on #germany
 GERMANY                 FAX: old fax number obsolete!!!
======================================================================

2. Deformable terrain

3. OPTICAL CHARACTER RECOGNITION QUESTION

4. .......How To Make Backgrounds.......

5. Optical Character Recognition Software

6. OpenGL Frequently Asked Questions [2/3]

7. OCR=Optical Character Recognition Software

8. Saving RLA depth data??

9. Optical Character Recognition software recommendations?

10. character recognition

11. Character Recognition Algorithm Needed

12. Hand printing character recognition

13. looking for character recognition software