For something like this I'd suggest you use HTML::TokeParser which has an excellent tutorial: HERE.

I think you'll find it to be far easier and probably somewhat more effeceint.

Hope that helps
Chris