Sure. Get the HTML with lwp, read the hrefs and srcs out of it with HTML::Parser (or friends), and then chop off the part after the last /. Lather, recruse, repeat. This, of course, wont catch directories that dont have anything in them used, but exist anyway, and directories that have only their index.html linked to with the implicit directory-name-without-trailing-slash redirect.

If you want somthing thats a little less of an ugly hack, hope you can use FTP to get the content.

PS -- sorry about the bad contractions, but the quote key on this keyboard is broken.