After processing, this method should return a set of bounding boxes of
regions within the image mapped to a String that contains the recognised
text. If the OCR processor cannot extract regions (it just reads whatever
image it's given), then the map can be singular with the rectangular
bounding box the same as the image bounding box.