A couple years ago I wrote an autocrop algorithm for the Internet Archive, but never had time to optimize it, and it was never deployed because it was too slow to run in realtime. The algorithm works well for books imaged using the archive.org Scribe bookscanner. Here is example output of the autocrop algorithm, with a comparison of the same book cropped and deskewed by a human.
The Internet Archive has launched a project to make books available to the print-disabled community in the DAISY format. Donate books to IA and we will turn them into DAISY books, which will be made available on the relaunched Open Library site. [more inside]