From: Brian Buhrow (
Date: Thu Aug 01 1996 - 07:08:39 PDT

I intend to try the methods Dave is using at the Visions Lab.

Dave writes:

Dear John,
        Oops, I forgot to tell you about our progress with scanning and
reverse tranlsation of hard copy Braille. Let me update everybody on
what we have done:

The problem of converting hard copy Braille is a very important one.
This problem first cam to our attention when our students started
wondering about how the math professor was going to grade the homework
that they did on the Perkins Brailler. Of course, for those students who
use a Braille-n-speak or something, the problem has already been dealt
with (you can easily reverse tranlsate literary Braille, and we are
almost done with our reverse Nemeth translator). But, there are a lot of
students who still like Braillewriters and many existing hardcopy
documents that need to be in electronic format.

To solve the homeowrk problem, we have the students always type on
triple-ply, carbonless-copy NCR paper. Then, they peel off the back
copy, keeping the top 2 copies. This leaves them with a good quality
Braille document, and us with an ink image of the Braille dots. We then
scan the ink dots into the computer and reverse translate. We have been
able to do this VERY ACCURATELY with a program called Character Eyes Pro,
from Ligature software. I have trained this program to recognize our
Braille font (it is a trainable neural net sytem, very nice). We can
thus take the scanned-in Braille and run it backwards through Duxbury or
something. Since we have our Braille font installed as a true-type font
in windows, it can be recognized by OCR just like any other font, with
some training of course.

Now, what happens when you have a hard copy already? We have tried
several techniques. First, you can rub carbon pencils over the paper,
scan in the image, take a photo-negative with Corel Draw, and then OCR
that file. As you can imagine, this is not very reliable. Furthermore,
it destorys your original (or at least makes it very messy to read!)

Another technique, then, is to use very strong sidelighting and Xerox the
Braille on highest contrast. This actually works very vell, but is very
time consuming, and can be expensive since you often have to Xerox the
document many times to magnify the contrast.

The approach that we have settled on is to Xerox the Braille once on high
contrast, and then scan it in. Using Corel, we are workling on writing
image processing routines to pull out the signal (the dots) from the
noise (the white background). This looks very promising, but will take
us some time to finish.

I know that other groups have worked on this problem. I think that, out
of necessity, we will have a good system worked out by the end of the
year. But, if anyone else has other ideas, I'd be happy to hear them.

Please poet this if you think it would be useful.
Warmest regards,

Dave Schleppenbach, VISIONS Lab Director
Purdue University Department of Chemsitry
1393 BRWN Box # 725
West Lafayette, IN 47907
(317) 496-2856 phone
(317) 494-0239 fax

This archive was generated by hypermail 2b29 : Sun Dec 02 2012 - 01:30:04 PST