Hello,
I've used Kurzweil 1000 to recognize a few PDF documents, though none that
were image based. It worked well. The only problem was that, as with any
OCR, there were recognition errors. But this does seem to be a viable way
of accessing PDF documents.
Regards,
Anna