I am assuming that the poster of the question calls
unrestricted books" the books that have no DRM, i.e., no technical
devices intended tp prevent some uses, such as copy, search, or
adaptation to new reading devices. All books are born legally restricted by
copyright, at least today, until they enter the public domain. Very few rightsholders fully remove copyright
restrictions by providing adequate public licences.
Though I have read many papers on the topic, and written some, I doubt there is such a
document as requested in the question. The reason lies in the
diversity of objectives of the numerisation projects, on the lack of
standard definition of what a book is, on the wide variations
encountered on book counts in various studies (notably those concerned
with assessing the proportion of orphan books in various libraries).
The definition of a book depends on a variety of parameters, including
content and purpose, commercialization, relevance of distinct
editions, choice of illustrator, each of which may impact some numbers
by an order of magnitude. Does the Bible count for one, or for one
hundred thousands? What about Shakespeare's works, or Alice in
Wonderland? Does a single poem count as a book? What about a short story, a music score, a ship log, a catalog, a birth registry? Recall that what figures may be available may have been
computed with different issues or purposes in mind, hence with different visions (often not explicited) of what counts as a book, and of what counts as two different books. Hence, rather
than pointing on a specific document as requested, I try below to give a direct
answer, based on figures I may remember or retrieve. But a short crisp
answer would be misleading given the very fuzziness of available
information, and the fact that the question itself is thus
unwittingly very imprecise.
The production of unrestricted books by Project Gutenberg and similar
projects is tiny compared to raw digitization and OCR, as donne by
Google and large public and academic projects, for example by national libraries. It is on the order of
100 000 books, a magnitude order of 1/1000th of the total
number of books, that seems to range somewhere between 50 and 130
millions, depending on how books are counted. Large digitization
projects like those led by Google, National Libraries and large
institutions are about two orders of magnitude more productive, some producing several millions of digitized books, though
exact numbers are sometimes hard to assess.
However, if you consider quality digitization with proofread text and
structured formatting like the epub standard, efforts like the
Project Gutenberg seem significantly more productive, as large
digitization projects produce only a few thousands, unrestricted,
quality e-books (from what I have been able to check).
This is not surprising since large digitization projects are intended
mostly for preservation purposes, and for accessibility by the
research community (or for indexation purposes in the case of Google), while Project Gutenberg and its likes are intended for availability/accessibility purposes, i.e. to give convenient access to existing literature to a wider reading public. Raw
digitization for conservation performed by large project is fairly
cheap and fast, while quality digitization require long and costly
human intervention (about one order of magnitude higher), that can be
well covered by crowd-proofreading and editing.
Another aspect is that, while the total number of books is very large,
only a small fraction is of interest to a wide audience. So the
relevant corpus for Project Gutenberg is probably much smaller than
the relevant corpus for large digitization project that have to deal
with all kinds of archives, inventories and the like. This is
compounded by the fact that the books that are of interest to a wide
audience are to a large extent still under copyright, hence much less
likely to be made available legally without restrictions, by any digitization project.
This actually stresses the importance in the book count of orphan
books, and more generally of books not cared for by rightholders, and
whether they can, or cannot, be made legally available, with or
without restrictions. This may represent on the order of half the
books in copyright. In the USA, it is an important aspect of the legal
battle around the Google Book Search Settlement Agreement, and it is
the object of a specific law in France.
So, I would tend to believe that cooperative projects like the
Project Gutenberg are the major source of unrestricted ebooks, if you
consider reading resources for a wide public. But, it makes little
sense to compare the output of two types of projects that have different
purposes, at least in the medium term.
Still, it should be noted that preservation is probably a more urgent
task than high quality digitization. It is rather sound that
preservation projects will avoid diverting much of their limited
resources on costly high quality output, other than for public
relation purposes. Also, the mass of documents
digitized by large projects provide a good source of raw material for
smaller projects (in size) that are aiming for quality output
that has a chance of attracting some readership.
We should also remember that digitization is not the only source of
ebooks. There are authors, publishers and institutions that produce a very significant number of
new books without DRM restrictions, though there may be legal
restrictions on use not enforced with DRM, as was the case before with
classical paper books.
Though I spent significant time analyzing the question and writing the
answer with some figures to substantiate it, it ended up being far too
long, and messed up by the heterogeneity of sources. Hence I decided
to cut it short here. If interested, you may read the rest in a
previous version from the history of this anwer, which contains some additional figures and references from various projects.
2 related questions:
What kinds of DRM are used in eBooks? Is it possible to purchase DRM free eBooks?
Some classic literature is available for free - but where?