tags:

views:

147

answers:

2

I have been tasked with finding an Imaging/OCR/ICR SDK that we can use in our Delphi Application.

I needs to do the following:

  • Work with Delphi XE (Don't care if it's ActiveX or a DLL)
  • It needs to work on a server with zero human intervention.
  • Needs to Support OCR and ICR in a given Rect or Zone.
  • Needs to return confidence level for a given zone and specific characters in the zone.

Cost is not currently important. We are willing to look at both Open Source and Commercial Solutions. We have used Tesseract but were not impressed by it's quality when compared to some desktop OCR products.

Does anyone know of a library that can meet these needs? If you have used that library what was your experience (bad experience reports are welcome)

+2  A: 

There are some options out there.

  1. http://www.rerecognition.com - Has been around for a long time and gives good results. This engine is used by most of the major suppliers in their products. Can read zones. Not as strong when reading full page. Delphi support.

  2. http://www.abbyy.com - FineReader SDK. Shouldn't be too hard to use with Delphi as it is an ActiveX. Good for zones and full page reading. Last time I looked it used and ActiveX inferface.

  3. http://www.captaris-dt.com/product/capture-recognition/features/en/ - Recostar. Has been around for a long time and was once the best solution. I haven't used it for a few years so I am not sure how easy it is to integrate with Delphi. Very powerful for reading zones. I am not sure about full page, you will need to check if you need this feature.

You won't find any open source OCR engines that can match any of the options listed above. All engines listed above perform Handprint and Machineprint. Previously I have integrated each of these engines into Borland C++ Builder.

You will need to write a wrapper around each of these engines to get the server functionality you require.

There is much more to an OCR/ICR solution than you may realise - image processing, scanning, correction etc... and it also depends on the type of documents you are looking to read.

Andrew Cash
You could also try Mitek although I have no experience with it. http://www.miteksystems.com/ImageNet_DataCapture_details.asp
Andrew Cash
A: 

If you're ok with calling an API on an external server, give WiseTrend a try ( http://www.wisetrend.com/wisetrend_ocr_cloud.shtml ) - it's a REST API for OCR with per-page pricing, based on the ABBYY engine (high recognition quality even for low-quality images)

Eugene Osovetsky
I have to OCR highly sensitive tax forms. They can not leave the building for processing.
Robert Love