Harness

Considering that font differences can be very common between GUI apps, we cannot assume that any text will always be pixel-for-pixel identical.

The idea here is to use optical character recognition (OCR) to scan any screen images and match the regions of text. These regions will need to be known so that can be handled differently in the image comparison code (see #2865). But we will also want to potentially actually check the text against lists of expected text.

Some early resources to check:

http://stackoverflow.com/questions/1813881/java-ocr-implementation
https://github.com/axet/lookup
http://roncemer.com/software-development/java-ocr/
http://sourceforge.net/projects/javaocr/