The 'funny symbols' also tend to be the fastest. It was
the fastest m// routine, but I found it strange that a s///
was still faster.

As mentioned the substr is probably the best anyone can do
for performance, but for readability I would probably go
with the replace routine.

use Benchmark;

$img = "http://www.my-website.be/gif/wwwtitle.gif";
timethese(1000000, {
    'regexp1' => sub { my($file) = $img =~ m#.*/(.*)# },
    'regexp2' => sub { $img =~ m#[^/]+$#; my $file=$& },
    'regexp3' => sub { $img =~ m#.*/#; my $file=$' },
    'replace' => sub { $file=$img; $file =~ s#^.*/## },
    'split'   => sub { my($file) = (split /\//, $img)[-1] },
    'substr'  => sub { my $file  = substr $img, rindex($img, '/') + 1 
+}
});
[download]

results:

Benchmark: timing 1000000 iterations of regexp1, regexp2, regexp3, rep
+lace, split, substr...
   regexp1: 26 wallclock secs (20.06 usr +  0.06 sys = 20.12 CPU)
   regexp2: 25 wallclock secs (20.02 usr +  0.06 sys = 20.08 CPU)
   regexp3: 17 wallclock secs (14.83 usr +  0.04 sys = 14.87 CPU)
   replace: 17 wallclock secs (13.05 usr +  0.03 sys = 13.08 CPU)
     split: 30 wallclock secs (24.56 usr +  0.09 sys = 24.65 CPU)
    substr: 11 wallclock secs ( 8.46 usr +  0.02 sys =  8.48 CPU)
[download]