Then opening the uncompressed.pdf in my favourite text editor, I found a block of text more than 50 lines long which I could see was obviously the code for the watermark.
I first used Philippe's solution to uncompress the PDF.
I needed to use a text editor to find and remove it. The watermark I needed to remove was a stream object (which is a multi-line block of code), not a single line, so a single line sed command wasn't going to work for me. Another add-on to Philippe's add-on to Dingo's answer.
