I guess if we are going to beat this thing to death, split() could also be used:

use strict;
use warnings;

my @images = qw{a.png b.gif c.svg d.blah.jpg ..};

foreach (@images) {
     my $after_last_dot = (split (/\./,$_))[-1];
     $after_last_dot //= '';
     print ".",$after_last_dot,"\n";
}

__END__
.png
.gif
.svg
.jpg
.     <=might want something else here?
[download]

I believe that the substr, rindex approach will be by far the fastest - these are very simple functions. The regex will be slower, but in my opinion, it is much easier to understand and I would prefer it for that reason. For most of my work, the speed difference would not be of any significance what-so-ever. There are of course always exceptions if you do something enough times! I suppose that split() performance would possibly wind up performance wise somewhere in-between? Although without benchmarking, I can't be sure. It could actually be slower than the first regex method because of more things pushed onto the output array.

Anyway in the spirit of "more than one way to do it", see split() solution. I did add code to handle the "undefined" case. The //= operator is a cool thing.