Speaking about the US: scans of public domain items are themselves in the public domain. The amount of work people have put into creating a copy does not enter into it: for a work to be copyright it needs to have originality. Which is a legal term that unfortunately means something completely different than in regular English.
Don't believe me, believe a court of law. The definitive case (again: for the US!) is
Bridgeman vs. Corel.
Claiming or suggesting otherwise by using harsh copyright disclaimers or watermarks is something called copyfraud. Unfortunately copyfraud is not illegal in most countries.
Although the principles underlying Bridgeman are valid in most countries, a lot of jurisdictions have exceptions that give a minor copyright (typically somewhere between 20 and 50 years after creation) to so-called sweat-of-the-brow work, that is: labour that does not require originality. Project Gutenberg works from the assumption that sweat of the brow does not create a copyright in the US, and so far it has managed to stay out of the clutches of copyright rent-seekers (but
not for want of trying).
Collections can have a compilation copyright in the US. This is because making the selection for a compilation is considered an act of authorship: making the selection has originality. However, this only prohibits you from copying the entire selection or recognizable chunks from it. Copying a single work from a collection of public domain works is not copyright infringement.
The Project Gutenberg license has nothing to do with compilation copyrights though: it is a trademark license. It prohibits you from distributing books as genuine Project Gutenberg books if you don't adhere to the license. Simply strip out the license and you're fine (funnily enough in most Project Gutenberg books, the license is the only copyrighted part).
IANAL, but I am a Project Gutenberg volunteer, and I can tell you that we would be in extremely hot water if scans of public domain works were copyrighted. Having said that, at Distributed Proofreaders we often adhere to reasonable requests by scan archives (for instance requests to be credited), simply because we want to maintain good working relations with these archives.
To answer vivaldirules' original question: if you are in the US, and the page scans are from a book published before 1923, the scans themselves are in the public domain.