Assuming that you know the hash name, and that it is a real hash, not a reference, this one would work.

Caveat! Your example has one key named "color" and the other named "colors". You should unify them.

#!/usr/bin/perl -w
use strict;
use Data::Dumper;

my $other = 'somescript.pl';
my $find = '%some_hash';
open SCRIPT, "< $other"
    or die "can't open\n";

my $first = "";
my $hash_text = "";
while (<SCRIPT>) {
    if (!$first) {
        $first =  /$find/;
    }
    next unless $first;
    s/^\s*my $find/$find/;
    $hash_text .= $_;
    last if /\)/;
}
close SCRIPT;
my %some_hash;  # must be the same name as the hash in the other scrip
+t
eval $hash_text;

my %colors ;

for ( keys %some_hash) {
    for my $color ( split /,\s?/, $some_hash{$_}{colors}) {
        push @{$colors{$color}}, $_;
    }
}

print Data::Dumper->Dump([\%colors], ['colors']),$/;
__END__
$colors = {
            'blue' => [
                         'MASSACHUSETTS',
                         'NEW_YORK'
                       ],
            'green' => [
                          'NEW_YORK'
                        ],
            'black' => [
                         'NEW_YORK'
                       ],
            'white' => [
                          'MASSACHUSETTS'
                        ],
            'red' => [
                       'MASSACHUSETTS'
                     ]
          };
[download]

Notice that this code will break if your hash contains a literal ")". But just to give you the idea.