What software would you recommend to upgrade your image to OCR (Optical Character Recognition)?

Question

We are currently exploring ways to improve image quality before sending to OCR. Currently, the OCR module uses the Scansoft API from Nuance (v15). We studied Lead Tools , but have since decided to look elsewhere. The licensing costs associated with Lead Tools are too high. To begin with, we are looking for simple image enhancement features such as articulation, stain removal, line removal, punch hole removal, sharpening, etc. We use a combination of .NET and Java software, but a Java solution is preferable.

Answer 1

Kofax is good for preprocessing, but for the types of cleanup you are talking about, it may be redundant if the images are not very bad. If your specialty is not in image processing, I would recommend working with a provider that performs image cleaning and OCR so that you can focus on the value that you really add.

We license the ABBYY OCR Developer Kit ( ABBY SDK ) and have found it to be excellent for both image processing and OCR. The API is quite extensive, and the sample applications, help, and support were impressive. I definitely recommend a look.

Answer 2

Disclaimer: I work for Atalasoft

We have these features and licensing without a license for .NET.

http://www.atalasoft.com/products/dotimage/

We also have OCR components, including the .NET wrapper for Abbyy, Tesseract and others, as well as for searching for PDF images (image over text in PDF format).

Answer 3

Not sure if this will meet the standards you need, but maybe you should take a look at some of the Paint.Net APIs. I don’t know how easy it would be to extract their image processing algorithms for use in your project, but I believe that they do some of the things you are looking for. In addition, it is an open source project with a MIT license, so it should be very convenient for use in business.

Answer 4

KOFAX VRS Research at KOFAX.com

Answer 5

Perhaps JMagick is the ImageMagick open source Java interface. It is implemented as a thin layer of Java Native Interface (JNI) in the ImageMagick API. It is licensed under the LGPL, so it should not be a problem license.

http://sourceforge.net/projects/jmagick/

Answer 6

I propose Intel for its licensing at zero cost at runtime.

Answer 7

Depends on the quantity and quality of the source images. Managed code sets and image sets will work, but this is not always the best solution if you want to process several million images. For small parties and tight budgets, I agree with previous posters that projects such as Aforge, Paint.NET and other open source libraries will do the trick. Of course, you are on your own, if the results do not improve ... At least this will allow you to put everything you need in one application for low cost.

If you process several hundred thousand images a month, I would suggest you divide the process into a smaller step in the workflow and configure each of them until your cost per image becomes as close to zero as possible. You will find that the OCR results first grow rapidly and then level out sooner than you expected. (I'm not a big fan of OCR, but he has his place)

I am using Recogniform's commercial Windows product to process and clean images before OCR in batch mode using scripts configured for different types of images. If the image does not work with QC or is rejected by the OCR mechanism, it is manually restored using a special .NET application created using the Atalasoft tool. A package handles everything and deals only with what fails.