The main wheels I would suggest would be HTML::Parser or any of several others in the HTML:: tree, and URI.

use URI;
use HTML::Parser;

my $parse = new HTML::Parser( default_h => [ sub {print shift}, 'text'
+ ],
  start_h => [
  sub {
    my ($tag, $attr, $origtext) = @_;
    if (($tag eq 'img') and ($attr->{src} =~ m#^(?:http://)?myserver\.
+#)) {
      $attr->{img} = URI->new($attr->{img})->abs('http://myserver.com'
+);
      print "<$tag " . join (' ', map {$_.'="' . $attr->{$_} . '"'} ke
+ys %$attr) . '>';
    }
    else {
      print $origtext;
    }
  }, "tagname, attr, text"]
);

$parse->parse(join('',<>));
[download]

However this code is untested and may well need some work. It currently would take input on STDIN and produce results on STDOUT.