$c = "a_c_code";
$c =~ s/(?<![^_])(.)/ucfirst($1)/ge;
print $c;
[download]

This ucfirsts any character that doesn't follow a non-_. (The double-negative condition is a more terse way of saying "follows a _ or at the beginning of a string": s/(?:\A|(?<=_))(.)/ucfirst($1)/ge.)

Why ucfirst instead of uc when only dealing with one character? Because there are ligature characters (the only latin1 instance being "\xdf") where the two produce different results.