18

I have a workflow whereby I scan paper documents into searchable PDFs using a Fujitsu ScanSnap S500 document scanner. I am not a great fan of the bundled software, but it's dead simple to use: place a stack of paper at the top, push the green button and a searchable PDF comes out.

Now, I'd like to do something similar on Linux (Ubuntu 10.10). The scanner is supported out of the box.

I've looked at gscan2pdf and XSane:

  • XSane looks powerful, but not really suitable as a workflow solution;
  • gscan2pdf is a bit closer to the "push the button, get the PDF" ideal, but still not 100% there.

Any other software you can recommend (free or otherwise)?

fixer1234
  • 27,064
  • 61
  • 75
  • 116
NPE
  • 1,229
  • 2
  • 11
  • 22
  • i use pdf-cups, but it is an image not searchable text – RobotHumans Dec 12 '10 at 18:08
  • 1
    What's 'not 100% there' with gscan2pdf? – digitxp Dec 12 '10 at 21:04
  • @digitxp I didn't want to clutter the question with a laundry list of issues, likes and dislikes for any product. However, since you ask, in `gscan2pdf` I had weird artefacts with 'unpaper'ing, the OCR was mostly unusable (some engines better than others) and overall it wasn't quite as streamlined as the original solution. Anyhow, the essence of my question is to see what else is out there so that I can try various solutions and see what works best for me. – NPE Dec 13 '10 at 09:13
  • @digitxp I've just re-read my previous comment and it sounds pretty negative. That was not the intention. `gscan2pdf` is actually fairly close to what I am looking for, but there are areas in which it is sadly lacking compared to the original solution. – NPE Dec 13 '10 at 09:20

1 Answers1

18

Here are some things that I found when researching this earlier this year. Sorry, I can't post more than one hyperlink due to my limited rating, so you'll have to Google for the links.

gscan2pdf

A really good GUI system that can use various OCR engines for the backend. This probably will meet your one-touch solution (and digitxp already mentioned it).

Tesseract OCR Engine

Can be used with gscan2pdf.

Ocropus

I didn't get very far with ocropus since it wasn't recognizing text without extensive training. It would probably be really good for books, but didn't work well for me with bills and such. YMMV.

Cuneiform

I had the best success with Cuneiform and was able to create searchable PDF's by scripting commands similar to the following workflow:

# extract images from scans
# (not shown)

# convert to black-and-white
optimize2bw -n -i nuance-test.png  -o bw.bmp                               

# do the OCR process and generate an hOCR file
cuneiform -l eng -f hocr -o nuance-test.html bw.bmp

# reassemble the original image with the hOCR file to generate a new PDF
hocr2pdf -s -i nuance-test.png -o nuance-test.hocr.pdf < nuance-test.html

You will also need to install the exactimage package.

Various open-source projects for OCR'ing PDF's use Cuniform and hocr2pdf as well:

  • WatchOCR
  • Archivista

Let me know what you find out!

akira
  • 61,009
  • 17
  • 135
  • 165
Eric
  • 306
  • 2
  • 3