I wonder if this could be used for preprocessing documents before doing OCR...