tags:

views:

582

answers:

6

I'm looking to add OCR and Barcode Recognition to an in-house application. Since none of the OpenSource offerings worked out for my problem I'm now looking into commercial OCR libraries. This seems to be a strange world with license fees per page processed, a "call us for a quote" approach to selling and everybody claiming he has the far superior engine.

Can anybody share experience with these libraries? Are they worth their money?

+1  A: 

Look at Abbyy SDK

Used it for food delivery application (OCR and barcodes), works quite well.

Quassnoi
While I have no problems with paying good money for good software I find the whole licensing approach at Abby distasteful: Dongle, enforced limit on the pages processed per month, lot's of add on costs (want to open a PDF? add 30%$! Want to save to PDF? Add 40%$). That kept me away from it so far.
mdorseif
+1  A: 

I've been using Aspose Barcode for about 6 months now to create and read QR barcodes on email attachments and faxes. It is pretty easy to set up and works consistently. A developer license will run about $399.

DavGarcia
+1  A: 

I've used ABBY for passport data and while I can't comment on the pricing, the results are first rate. Even in non-English (Cyrillic) OCR.

kingmunkyman
+1  A: 

As far as Open Source the options are Tesseract, Cuneiform or one of a few other options. So far I've not found any of them to be accurate enough for my needs (lack of layout analysis primarily). I've been looking at commercial solutions and agree the per-licensing is annoying. Nuance and Abby seem to be the biggest players and they're both expensive and limiting. ExactCode has some open source tools including barcode detection. The actual barcode part is commercial(?) but if you have it their code can compile it in and provide an API to it.

wdingus
A: 

If you just need the raw text (and maybe character x,y), and you're on Windows, I recommend Transym's TOCR engine. It's a very developer-oriented product and company, just dedicated to turning images into text (like, char*) quickly and accurately. They do want to sell you a license per workstation, but also offer to negotiate multicopy licenses - I would guess for an in-house app it might even be something reasonable...

I've had dozens of my customers use TOCR over the last few years, and generally they love it.

But I can't wait for the day when OCR engines are licensed like PDF libraries. Or better, like LibTIFF and libjpeg.

Spike0xff