Maybe not a good answer, but here's how I've approached the problem in the past: With Adobe Acrobat 5, you can print to file the .pdf, so it's saved off as a Postscript file.

I have perl scripts which parse through (scanning page headers and footers), and pluck out the ones I care about. I guess it really depends on how comfortable you are with Postscript.

So you run your script against the .ps, generate a new .ps, and with Adobe Distiller you can convert that back to a .pdf. A longer process, but at least your input file is all ASCII, and you can unleash Perl's capabilities on it.