I assume that your problem lies within the regular expression and not in the CGI-handling, although your question dosen't really make that clear.

I'm taking this regular expression ("RE") from an old shell script called sitemap which did a site map in bash and some GNU tools - you might want to read more on regular expressions in perlre...

  # for offline conversion, set this to 1
  $offline = 0;

  undef $/;
  open( FILE, "< $filename" ) or die "Can't open $filename : $!\n";
  $HTML = <FILE>;
  close FILE;

  # Note that you can use \2 (or $2) in the new title part
  # to reference the old title.
  $HTML =~ s!(.*?<TITLE>)(.*?)(</TITLE>.*)!... new title here ...!ims;

  ($Left, $Title, $Right) = ($1,$2,$3);
  # Here, $Left contains everything left of the page title
  # $Right contains everything right of the page title
  # and $Title contains the page title
  # here you could do some more magic to the page title
  # before assigning it back to $HTML ...


  if ($offline) {
    rename $filename, "$filename.backup";

    open( FILE, "> $filename" ) or die "Can't create $filename : $!\n"
+;
    print FILE $HTML;
    close FILE;
  } else {
    print FILE;
  };
[download]

As you might (or might not) notice, this code will have problems with malformed HTML or a document that contains two TITLE tags, only the first such tag will get handled. Also, if you have a commented-out TITLE tag, this will also confuse the RE. But as you say that you want to use this in a template, you can craft your template around that, I hope :)