OCR program?

[nifty542]

Hi, all

I'm almsot there! There, as in ready to ditch Windows altogether.
I have a very specific problem- as a teacher, I sometimes need to scan in documents for editing.
So far, I haven't been able to get an OCR program to work in Linux, or at least, not accurately enough.
Could anyone please recommend one? I am quite happy to use the command line, but can't find a detailed "how to" for an OCR program.
Regards, and thanks in advance

Nifty

[Matt__]

Abbeyyocr seems to do a good job if this blog is to believed, but it is a commercial product.
10 Euros/9 GB pounds. There is a free 15 day trial here.

[nifty542]

Hi

Thanks for the reply. I don't mind paying, provided the product works. I'm still wondering about a "how to", though. Do you start with Simple Scan?
Regards
Nifty

PS. What have they done to the thanks button?

[Matt__]

ischool tutorials

[nifty542]

Hi, Matt

Thanks for that. I shall give this a try now.

Your halp is much appreciated.

Regards

Nifty

[ajgreeny]

I use tesseract v3 in Lucid 10.04, which can be added from a number of ppa repositories and is a big jump forward in accuracy terms, in my opinion.

I can get it to use other image formats than the black and white image.tif images that are required for tesseract v2, so I suggest you at least have a look at that before spending any money, as you may find that it is good enough for what you want.

I use xsane for the scan process when needed, though I see no reason why simple scan would not work just as well; it's just that I've never tried it but will now do so.

Yes, just tried and a full page of Times New Roman, 12 point text, scanned with Simple-scan at 300dpi and saved as an image.png or an image.jpg file has just OCR'd with only a tiny error of an added ' at one point where the page was slightly mucky. Both jpg and png files worked with no problem.

[forrestcupp]

There's no reason to go beyond the free software included in the repos. Open up Ubuntu Software Center and search for gscan2pdf. You can scan multiple pages and save as a pdf or tif or something, and it seems to have decent OCR capabilities.

It's the best scanning/OCR software I've seen on here yet.

[ajgreeny]

There's no reason to go beyond the free software included in the repos. Open up Ubuntu Software Center and search for gscan2pdf. You can scan multiple pages and save as a pdf or tif or something, and it seems to have decent OCR capabilities.

It's the best scanning/OCR software I've seen on here yet.

Not much help if you want to get a text document that you can edit to change the text, surely? Or does it allow you to select text from the pdf it makes and then copy/paste that to a document?

When I use OCR it is almost always because there is a need to make considerable edits to all the text; to make a pdf out of the sheet of text does not seem to help much, or have I totally misunderstood what gscan2pdf does?

[forrestcupp]

Not much help if you want to get a text document that you can edit to change the text, surely? Or does it allow you to select text from the pdf it makes and then copy/paste that to a document?

When I use OCR it is almost always because there is a need to make considerable edits to all the text; to make a pdf out of the sheet of text does not seem to help much, or have I totally misunderstood what gscan2pdf does?

For the OCR feature, you copy the text and paste it into LibreOffice or something.

Edit: I just tried out the OCR on a screen shot I took of a Word document. You can't get a much better source than that. The OCR output royally sucked. Even the best OCR engine out of the 3 only got 2 words right. So this software is great for scanning multi-page documents, but not so great for OCR. The document was all in italics, but a decent OCR should be able to read a perfect image with italics.

[ajgreeny]

A screenshot will only be at a max of 96dpi, not the 300dpi needed for good OCR. Did you also use tesseract 3, or the v2 from the repos, as v3 is a lot better than v2.

I must try an italic page of print to see if I can do better than you managed. I'll report back asap.