I agree, you really don't want a regex. :-)

Here's my go using HTML::TokeParser::Simple.

#!C:/Perl/bin/perl.exe

use strict;
use warnings;
use HTML::TokeParser::Simple;

my $html = q{
  <a href="some.html"><img src="an_image.jpg"/></a>
};

my $p = HTML::TokeParser::Simple->new(\$html);

my $in_anchor;
while (my $t = $p->get_token){
  if ($t->is_start_tag('a')){
    $in_anchor++;
    next;
  }
  if ($t->is_start_tag('img') and $in_anchor){
    my $src = $t->get_attr('src');
    print "$src\n";
    $in_anchor = 0;
  }
}
[download]

output:

an_image.jpg
[download]