But, regexes might not be the best for matching this sort of thing, and actually parsing the HTML may be easier to maintain in the long run.

Yep, there's also been a recent CUFP which if nothing else has the merit of showing how to do that easily with HTML::SimpleLinkExtor.