[parkher]

So far I was not able to find anything better than PerfectEpub extension for OpenOffice.
Therefore I do this:

OCR in FineReader -> save as odt -> open in OpenOffice -> run PerfectEpub (after possibly other cleaning with regex find/replace, etc.) -> writer2ePub (or save as odt or as html and then use Calibre converter - whichever works better) -> SIGIL (where you again can do regex find/replace, merge/split if necessary, etc.)

However it is better to get rid of any page numbers / headers before PerfectEpub.
FineReader 11 is pretty good at recognizing headers/footers so they are not much of a problem.
PerfectEpub joins wrongly split lines (paragraphs) with one click and also splits wrongly joined lines, etc. I don't understand why FineReader can't do this itself, though. If it can, I need to find out how...

I use PerfectEpub on already made epubs and other formats too, if they have wrongly split lines or wrongly joined lines in them.
For an epub, I do this: epub -> htmlz -> extract files -> open in OpenOffice -> run PerfectEpub -> save back to html (or run writer2ePub)

The line joining / splitting in such cases when the information about the original pages is no longer available can be done with regex find/replace in sigil directly, but it requires multiple regex expressions to be used and different for pretty much each epub, so PerfectEpub is a much quicker solution.