Hello everyone,
I'm working on a project that requires text to be recognized from on screen applications running as flash player or similar. These are casino and / or poker tables.
They typically not only do not have windows controls that can be used to get text from them but also resist any OCR packages available in the first page of google search for OCR.
I have however found a project that successfully solves this problem using custum built CTransforms for text, images, hashes etc. The whole project is available in source code here:
http://code.google.com/p/openholdembot/source/browse/
Using OpenScrape one can easily connect to any window, get a screenshot, then create regions that require scraping, designate coordinates, foreground text color and color radius (all required for detection of text in that particular region). From there OpenScrape will return pixels, and next step is to create fonts (or manually identify specific characters if you want), which are later used to recognize text. The whole process is relatively simple and straightforward, and results are as accurate as the fonts / characters base you build using the tool. This can typically be brought to 100% as soon as all characters are captured using OpenScrape.
Now, the problem is that these functions are hardcoded inside OpenHoldem or OpenScrape and can not be easily moved outside those programs. Full source is available for download at above link.
I was wondering if there are people here who are skilled with C++ and interested in decoupling this powerfull OCR component from OH/OS and creating a sort of standalone package, DLL or similar, that can be integrated in other packages / applications with relative ease.
If you would like to explore this in more detail, please contact me at registered email.
Thanks all!
