I have a friend who's blind, and I have a PDF which I would like to give him to read.
The PDF is about 200 pages of beautifully-designed text (with a few pictures and figures containing text, but those are rare enough to be picked up by manual cleanup). Unfortunately, the logical structure of the PDF is represented badly: The PDF is unaware of its two-column text-flow, and none of the index and table-of-content entries are actually links.
I have easy access to a Linux machine and slightly-less-easy access to a Windows XP machine, and I know my way around regular expressions and scripting languages for automatizing post procession.
So far I have found a way to cut the PDF in half vertically (using code from http://snipplr.com/view/18924/split-crop-double-page-pdfs-in-two/, after converting the PDF to PDF1.4 so as to not contain a crossrefstream or whatever it was called), such that most content is in the right order, and then using pdftohtml to extract the text with some markup. Unfortunately, this is not able to reproduce the logical structure of the document (it recognizes some italic markings, but loses already all chapter, section and subsection headlines, which are always highlighted using consistent use of font size and colour, not to mention paragraph breaks etc.)
I had high hopes in Calibre, but that conversion tool cannot deal with badly-markedup two-column pdfs either, and also cannot derive structure from the fonts used, although it has some advantages for keeping paragraphs together.
How do I convert my PDF into a format suitable for access by a blind person?