Enhance an image to perform an accurate OCR recognition

[reck...@gmail.com]

Enhance an image to perform OCR recognition

I need to make a Java application that must extract relevant information from an image. the problem is that when i use available libraries like Tesserarct, AspriseOcr ... the results are not accurate.

I tell myself that it may be due to the following factor:

1 - The image contains different fonts
2 - some are in bold and some are not
3 - The font used to print the numbers is really not very difficult to recognize by the APIs that I use (I guess). This is a real problem because the goal is to retrieve information such as the invoice amount, the fee, the expiration date, the subscriber number, the subscriber name, the customer, the meter number , the period of consumption ...

4 - the last problem is that the system used for the printing of the character does not do it at the same places for two different bills.

My real question is if I can use Marvin with Java to craft the image before OCR recognition or how can I otherwise achieve my results?

see attachments for more details.

[Jhane Monares]

if you are making web app I think you can use this. http://jsfiddle.net/Tamersoul/dr7Dw/

--
You received this message because you are subscribed to the Google Groups "Marvin Project" group.
To unsubscribe from this group and stop receiving emails from it, send an email to marvin-projec...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[reck...@gmail.com]

Thank you for your answer, Jhane.

But when I visit the link that you provided I see three fields with HTML, JavaScript and CSS and I do not know what it's used for. I also see a button to import a file and another also to execute. I tried everything but I understood nothing at all ... can you help me?