Scanning pages will give you a HUGE book: megabytes. I dunno what is possible at your side, but I'll tell you what I did.
I had a PDF of a computer-related book of some 500 pages. With NITRO I made an HTML file of that. Then with a flat-text-editior, I stripped out all the unnecessary formatting (and that was A LOT!). Then I diveded the file into sections, corresponding to the Sections of the book. Thus I had some 18 html files. Then I made <H> tags for all the chapters and paragraphs, corresponding to the TOC structure.
With SNAG-IT, I took out the pictures and inserted them in the HTML files.
From the resulting html files I generated the NCX end .OPF file. I built an ANT script and a Perl script to do this automatically, but hand-work is possible.
The fact, that the book was all in HTML means, you can easily format it like any web page: chapres, headings, bullet lists etc.. all in HTML.
Then I zipped up the HTML and CSS files et voila: an epub.
Mind you: IT IS A LOT OF WORK , making the html's!
If you have access to dreamweaver, you could use that to do your lay-out, i guess. But HTML is the way to go...