A simple regex that seems to always work for me for pulling the 'real' address out is this:
$from = $1 if $from =~ /<(\S+)>/;
If $from was "Usern name" firstname.lastname@example.org, you'd wind up with 'email@example.com'. If $from was just 'firstname.lastname@example.org', it still would be since the regex wouldn't match.
As far as the problems of folding and multiple addresses in a field that jlongino pointed out, I think MIME::Head does a good job of taking care of the messiness of that for you. I have a script that used to do a lot of that manually, but I was always tweaking it for special cases that I'd never thought of. For example, addresses like "Name, User" <user@whatever> threw off my parsing of multiples because of the commas. And I started out thinking I only had to handle folding (wrapping onto multiple lines) for Subjects, but have now seen it in every header you mentioned, even the From header (The person had a very long user name in quotes). The perldoc is very long, but take a look at the 'get' and the 'unfold' methods to handle the problems of getting multiples and folding.