I have a large number of scanned pdf books. Manually entering the ISBNs is a pain so I made a change to the plugin to OCR the PDF to get the ISBN. I am on Linux and used the tesseract package to perform the OCR.
The processing time per book is anything between 3 and 30 seconds. This is fine for my purpose as I just select a bunch of books, hit 'Extract ISBN' and get a cup of coffee. It is still quicker than entering the ISBNs by hand. Interestingly for the one book that took 30 seconds, running the pdftoppm and tesseract commands from the command line for the first 3 pages (ISBN is on the 3rd page) only took 13 seconds. I have not investigated why it takes longer inside the plugin.
I am attaching the patch file here in case anyone is interested. It works for me but no guarantee it will work for anyone else as it is just hacked together and only works on Linux.
ocr_pdf.zip