That's a nice solution, but but I think it is missing a couple features: it doesn't catch the first word in the string; it doesn't lowercase the rest of each word; and it doesn't allow variable-width delimiters. Here's a modified version that addresses those issues.

sub ucwords {
  my($new, $delim) = @_;
  $new =~ s/($delim|^)(.*?)(?=$delim|$)/$1\u\L$2/sg;
  $new;
}
[download]

To match an entire word, from one delimiter to the next, I used a non-greedy match and a positive lookahead.