It's not easy, and there aren't any guides to the specific task at hand. Here's my educated guess as to what they've done (OMCDL).
1. Obtained copies of the Kalmus or Lucks reprints of titles for about 99% of what they offer. (This is very logical on their part. Kalmus has already done the copyright research, after all. Also note that the OMCDL files are all 10 x 13 pages. 10 x 13 is Kalmus' standard size for their regular orchestra series. Lucks uses this size as well. Breitkopf and other European publishers use a different size - approx. 9.5 x 12.25.)
2. Scanned this collection of parts at 300 dpi monochrome, deleting all titles and page numbers. (Kalmus and Lucks, while careful to get rid of all plate numbers and other obvious publisher info on items they reprint, seldom bother to mess with the titles and almost never add new page numbers - especially on orchestral parts where layout is critical for page turns.) Those original titles and page numbers are often a great help in publisher identification, BTW.
3. OMCDL then add in new titles, page numbers, headers (repeating the title at the top of every page), and the trademarked logo.
The Kalmus catalog is about 35-40 percent made of of reprints from the Breitkopf und Härtel Orchester-Bibliothek. This is especially true of the major core repertoire (Beethoven, Bach, Brahms, Mozart, Mendelssohn, Schumann, etc.). Other composers, like Debussy and Ravel, are easy to guess because they were only issued by one publisher (usually Durand in their cases) who has been reprinted by Kalmus, Lucks, Dover, etc.