17 Jan
2012
17 Jan
'12
8:50 a.m.
Hi Everyone,
I have a bunch of PDF files that contain the microcode listings for an IBM
7201-02 CE (enhanced system/360 model 65), like this one:
http://ibm360-console.wikispaces.com/file/view/QZ001.pdf. I need their
contents for the emulator that drives my '65 control panel. Unfortunately,
the OCR software I have tries to recognize English words, and makes
gibberish out of them. I'm only interested in the 1's and 0's, so it would
be wonderful if there was OCR software that you can tell only to look for
0's and 1's (or have some bias towards recognizing characters as a 1 or 0.
Is anyone here aware of such software, or can anyone recommend a program
that might do a good job with these? I've been copying sections by hand, but
that gets very tedious when there's 2,816 lines of data, each having 100 0's
and 1's...
Cheers,
Camiel
17 Jan
17 Jan
9:05 a.m.
I have a bunch of PDF files that contain the microcode
listings for an IBM
7201-02 CE (enhanced system/360 model 65), like this one:
http://ibm360-console.wikispaces.com/file/view/QZ001.pdf. I need their
contents for the emulator that drives my '65 control panel. Unfortunately,
the OCR software I have tries to recognize English words, and makes
gibberish out of them. I'm only interested in the 1's and 0's, so it would
be wonderful if there was OCR software that you can tell only to look for
0's and 1's (or have some bias towards recognizing characters as a 1 or 0.
Is anyone here aware of such software, or can anyone recommend a program
that might do a good job with these? I've been copying sections by hand,
but
that gets very tedious when there's 2,816 lines of data, each having 100
0's
and 1's...
Dunno about OCRs, but I believe a simple program can address this. If
the limit is 1 or 0 a simple OCR program can be coded, separing the chars
and looking for the histogram of each char...
1:34 p.m.
we were able to locate arrays like this using an autocorrelation function.
In python, which was simple read each page into a binary array. You
would have a sample 1 or 0 shape per what Alexandre suggests as well.
The autocorrelation function matches the shapes essentially. The
coordinates of the corners will be apparent after the correlation. I
believe you could find the 0's with one pass and 1's with a second one.
This will take a long time to run at full resolution, so you'd need to
break it up into small chunks, since correlation time increases
geometrically (I think that is the correct math progression). We were
able to find circles in this fashion on our binary arrays.
Jim
On 1/17/2012 6:05 AM, Alexandre Souza - Listas wrote:
I have a bunch
of PDF files that contain the microcode listings for
an IBM
7201-02 CE (enhanced system/360 model 65), like this one:
http://ibm360-console.wikispaces.com/file/view/QZ001.pdf. I need their
contents for the emulator that drives my '65 control panel.
Unfortunately,
the OCR software I have tries to recognize English words, and makes
gibberish out of them. I'm only interested in the 1's and 0's, so it
would
be wonderful if there was OCR software that you can tell only to look
for
0's and 1's (or have some bias towards recognizing characters as a 1
or 0.
Is anyone here aware of such software, or can anyone recommend a program
that might do a good job with these? I've been copying sections by
hand, but
that gets very tedious when there's 2,816 lines of data, each having
100 0's
and 1's...
Dunno about OCRs, but I believe a simple program can address this.
If the limit is 1 or 0 a simple OCR program can be coded, separing the
chars and looking for the histogram of each char...
2:06 p.m.
jim s wrote:
we were able to locate arrays like this using an
autocorrelation function.
In python, which was simple read each page into a binary array. You
would have a sample 1 or 0 shape per what Alexandre suggests as well.
The autocorrelation function matches the shapes essentially.
I don't know if you even need to do that, depending on the font style - the
ratio of dark to light pixels in a given 'cell' might be enough to
determine if it's a 1 or 0. This sort of approach does assume that the
data's in a nice enough format (e.g. fixed width font, corrected for any
page distortion, little "outside" contamination) to work, though.
Of course it also doesn't help the OP, who was looking for an existing
program to do it ;)
(And I don't know if any OCR process can be fully trusted, so it still
needs some form of human validation, or retention of the original scans
alongside the OCR data)
cheers
Jules
12:10 p.m.
On 17 January 2012 08:50, Camiel Vanderhoeven <iamcamiel at gmail.com> wrote:
Hi Everyone,
I have a bunch of PDF files that contain the microcode listings for an IBM
7201-02 CE (enhanced system/360 model 65), like this one:
http://ibm360-console.wikispaces.com/file/view/QZ001.pdf. I need their
contents for the emulator that drives my '65 control panel. Unfortunately,
the OCR software I have tries to recognize English words, and makes
gibberish out of them. I'm only interested in the 1's and 0's, so it would
be wonderful if there was OCR software that you can tell only to look for
0's and 1's (or have some bias towards recognizing characters as a 1 or 0.
Is anyone here aware of such software, or can anyone recommend a program
that might do a good job with these?
I am not sure how helpful this answer will be but Tesseract
(originally a commercial HP product, now in Google Code) has training
files for different languages. I have never modified them but my
guess, from looking at the documentation, is that you could make
training files for a binary language.
N.
18 Jan
18 Jan
4:26 a.m.
If you have the pro version of Acrobat it does this quite well from the PDF
but be sure to select the ClearScan option - its remarkably good compared to
traditional OCR methods.
++++++++++
Kevin Parker
Advanced Imaging
e: webmaster at advancedimaging.com.au
w: www.advancedimaging.com.au
m: 0418 815 527
++++++++++
http://au.linkedin.com/in/krparker
-----Original Message-----
From: cctalk-bounces at classiccmp.org [mailto:cctalk-bounces at classiccmp.org]
On Behalf Of Camiel Vanderhoeven
Sent: Wednesday, 18 January 2012 00:51
To: cctech at classiccmp.org
Subject: OCR software for numeric (binary) data
Hi Everyone,
I have a bunch of PDF files that contain the microcode listings for an IBM
7201-02 CE (enhanced system/360 model 65), like this one:
http://ibm360-console.wikispaces.com/file/view/QZ001.pdf. I need their
contents for the emulator that drives my '65 control panel. Unfortunately,
the OCR software I have tries to recognize English words, and makes
gibberish out of them. I'm only interested in the 1's and 0's, so it would
be wonderful if there was OCR software that you can tell only to look for
0's and 1's (or have some bias towards recognizing characters as a 1 or 0.
Is anyone here aware of such software, or can anyone recommend a program
that might do a good job with these? I've been copying sections by hand, but
that gets very tedious when there's 2,816 lines of data, each having 100 0's
and 1's...
Cheers,
Camiel
6:24 a.m.
If you have the pro version of Acrobat it does this
quite well from the
PDF
but be sure to select the ClearScan option - its
remarkably good compared
to traditional OCR methods.
Thanks to everyone who has replied with suggestions. Another suggestion I
got is to try Omnipage, so I got myself a 15-day trial version of Omnipage
Pro. It allows you to draw a rectangle on the page, then select "Numeric
data". It will then only perform OCR on that part of the page, and will try
to recognize everything as a number. When in doubt, it will bring up a
dialog with a detailed view of the problematic page so you can make manual
corrections. Combined with a bit of manual work to fix things up, this seems
to work fairly well.
Cheers,
Camiel
19 Jan
19 Jan
9:36 a.m.
On 1/18/2012 6:24 AM, Camiel Vanderhoeven wrote:
Thanks to everyone who has replied with suggestions.
Another suggestion I
got is to try Omnipage, so I got myself a 15-day trial version of Omnipage
Pro. It allows you to draw a rectangle on the page, then select "Numeric
data". It will then only perform OCR on that part of the page, and will try
to recognize everything as a number. When in doubt, it will bring up a
dialog with a detailed view of the problematic page so you can make manual
corrections. Combined with a bit of manual work to fix things up, this seems
to work fairly well.
I realize you may have already got your answer, but abbyy finereader pro
works very well.
It also has an option for numeric only data, including the ability to
limit which numeric characters to include(ie, only 0 & 1, if you
wanted). You can also train it to do pattern recognition, if the
recognition of the font doesn't work naturally well.
With most of these programs, the quality of your original source
material seems to be the determining factor on how much manual work
you'll need to do afterwards to correct errors.
BTW: what's the end format?
http://www.abbyy.com/
I'm unrelated to company......but it does sound like a neat problem.
Cheers,
Camiel
Good luck,
Keith
11:44 a.m.
In article <4F182A87.8010800 at verizon.net>,
Keith M <keithvz at verizon.net> writes:
With most of these programs, the quality of your
original source
material seems to be the determining factor on how much manual work
you'll need to do afterwards to correct errors.
Yep.
I've found that doing things like eliminating speckling (i.e. stray
single black pixels) from the input will do wonders for the quality of
OCR output.
--
"The Direct3D Graphics Pipeline" -- DirectX 9 version available for download
<http://legalizeadulthood.wordpress.com/the-direct3d-graphics-pipeline/>
Legalize Adulthood! <http://legalizeadulthood.wordpress.com>
4602
days inactive
4604
days old
8 comments
8 participants
participants (8)
-
cym224＠gmail.com -
iamcamiel＠gmail.com -
jules.richardson99＠gmail.com -
jws＠jwsss.com -
keithvz＠verizon.net -
legalize＠xmission.com -
pu1bzz.listas＠gmail.com -
trash80＠internode.on.net