home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
OS/2 Shareware BBS: 22 gnu
/
22-gnu.zip
/
gocr035b.zip
/
README
< prev
next >
Wrap
Text File
|
2002-02-03
|
6KB
|
143 lines
GOCR/JOCR
GOCR is an optical character recognition program, released under the GNU
General Public License. It reads images in many formats (pnm, pbm, pgm, ppm,
some pcx and tga image files (or PNM from stdin); if pnm-tools installed and
running linux-like system you can also use pnm.gz, pnm.bz2, png, jpg, tiff,
gif, bmp and others) and outputs a text file.
To see installation instructions, see the INSTALL file.
How to start? (QUICK START)
---------------------------
You'll probably want to use one of the frontends available, such as the TCL/TK
or the GTK. They make your life much easier.
Some examples of how you can use gocr:
gocr -h # help
gocr file.pbm # minimum options
gocr -v 1 file.pbm >out.txt 2>out.log # generate text- and log file
djpeg -pnm -gray text.jpg | gocr - # using JPEG-files
gzip -cd text.pbm.gz | gocr - # using gzipped PBM-files
giftopnm text.gif | gocr - # using GIF-files
gocr -v 1 -v 32 -m 4 file.pbm # zoning and out30.bmp output
xli -geometry 400x400 out30.bmp # see details using xli (recommanded viewer)
wish gocr.tcl # X11-tcl/tk-frontend (development version)
How to get image files?
-----------------------
Scan text pages and save it as PGM/PBM file. Use a program such as The GIMP.
You can also use PNM-tools to convert several image formats into pbm/pgm.
djpeg can be used to convert jpeg into pgm
djpeg -grayscale -outfile file.pgm infile.jpg
Generate our own using TeX+DVIPS+GS or other programs, or convert existing
PostScript or PDF files to an image format. Generate provided examples:
make examples.
Memory limitations
------------------
WARNING!!!
If you use a 300dpi scan of A4 letter, the image is about 2500x3500 pixels and
gocr requires 8.75Mb for storing the picture into the memory. Not only that,
but gocr may create a 2nd copy, using a total of 17Mb. This is independent
of using b/w or gray-scale images. Be sure that you have enough RAM installed
in your machine! Alternatively you can cut the picture into small pieces.
You can use the pnmcut, from the netpbm package to cut the file. Example:
pnmcut -left 0 -right 2500 -top 0 -height 1000 bigfile.pnm > smallfile.pnm
And then use gocr in the cropped image as usual. Take care: if you chop the
characters, gocr won't be able to understand that line.
Future versions will take care of this issue automatically.
Limitations
-----------
gocr is still in its early stages. Your images should fit in these requirements
if you want a good output:
- good scans (all chars well seperated, one column, no tables etc, 12pt 300dpi)
should work well
- fonts 20-60 pixels ( 5pt * 1in/72pt * 300 dpi = 20 dots )
- output of image file for controlling detection
And note that speed is very slow (this will be changed when recognition works
well)
12pt 300dpi 1700x950 16lines 700chars 22x28 P90=40s..90s v0.2.3 (gcc -O0)
You can try to optimize the results:
- make good scans/treat image
- try to change the critical gray level (option -l <n>)
- control the result on out30.bmp (option -v 32)
- enlarge option -d <n> for high resolution images which are noisy
- try different combinations for option -m <n>
- for thousends of documents with same font
you can use/create a database (-m 2/-m 130)
What does >> NOT << work at the moment:
- complex layouts (try option -m 4)
- bad scans, noisy/snowy images, FAX-quality images
- serif fonts, italic fonts, slanted fonts
- handwritten texts (this is valid for the next ten years)
- rotated images (but slightly rotated images should be no problem)
- small fonts (fax like) or mix of different font size
- colored images (use gray or black/white)
- Chinese, Arabian, Egyptian, Cyrillic or Klingon fonts
- using database (create_db is for developer tests)
How it works or how it should work?
- put the entire file into RAM (300dpi grayscale recommended)
- remove dust and snow
- detect small angle (lines which are not horizontal)
- detect text boxes (option -m 4)
- detect text-lines
- detect characters
- first step recognition (every character has its own empirical procedure)
- no neural network or similar general algorithms
- analyze not detected chars by comparison with detected ones
- try to divide overlapping chars
- testwise: compare all letters (like compression of pictures)
- for more details look to the ocr.tex documentation
How can you help gocr?
----------------------
- Send comments, ideas and patches (diff -ru gocr_original/ gocr_changed/).
- If you have a lot of money, spend a bit (www.paypal.com).
- I always need example files (.pbm.gz or jpeg <100kB) for testing
the behavior of the ocr engine under different conditions,
because scanning does take a lot of time which I do not have.
But do not send files which are not convertable by commercial ocr programs
or which are protected from copying and electronic processing by copyright.
That will help, to get the world's best OCR open source program. :) Thanks!
- Send me your results (errors,num_chars,dpi) and if possible results
and name of professional OCR programs for statistics.
- Read OCR literature, extract the essentials and send a short report
to me ;).
- If you have a good idea, how to manage some OCR-tasks, tell me!
- Tell your friends about gocr. Tell me about your success. Be happy.
After all, is it gocr or jocr?
------------------------------
The original name of this project is gocr, from GNU Optical Character
Recognition. Another project is using the same name, however; so the
name was changed to jocr. If you have a good idea for a name, please
send it.
Latest news
------------
http://altmark.nat.uni-magdeburg.de/~jschulen/ocr/index.html
http://jocr.sourceforge.net
Authors:
jschulen@gmx.de (Joerg)
brunobg@geocities.com