Check out tesseract, it should do the job: https://code.google.com/p/tesseract-ocr/wiki/ReadMe
You can run:
tesseract Sometext_big.png out.txt
And out.txt will contain "Some text here".
Then it's just a matter of some shell scripting to call find
to find all of the images of a particular type, run them through tesseract and see if the output file contains the text you want.