Your never going to be able to tell a person's name unless they explicitly tell you in some way that is their name, even if implied with nothing more than formatting. What if the picture was called the "Mona Lisa"? You're going to have get down and dirty with an html parser, try HTML::TokeParser::Simple, remember you get access to HTML::TokeParser through it, and I belieive HTMl::PullParser. Block out the tags you know you won't need, which is probably any non-table tag, non-image tag.