Here's something you may want to check out..

http://www.cs.berkeley.edu/~phelps/Multivalent/index.html

http://www.cs.berkeley.edu/~phelps/Multivalent/Tools/pdf/Uncompress.html

These tools are written in Java and require JRE 1.4.1
I haven't tried any of them yet, but should be easy to implement with system calls or possibly Inline::Java

From the documentation:

Uncompress for Hand Editing or Examination

The written file leaves content streams uncompressed and available for inspection or hand editing. With reference to Adobe's PDF Reference, available online, you can arbitrarily change the PDF, anything from correcting bad OCR, to fixing typos on pages without having the generating application (text is not reflowed), to adding title and keywords, to authoring annotations to diagnosing problems.

Uncompressed content streams are pretty printed to better show structure and objects are labelled with the page numbers they're used on. For Western languages, it is straightforward to identify the character strings and edit them. One must be careful to edit with a text editor, such as Emacs, that can handle binary data and does not translate unfamiliar characters or line endings.

The edited file written by the text editor should be passed through the Compress tool to recompress the streams and rebuild the cross-reference table.

Hopefully someone will find these useful