It might be faster and more accurate if $count was derived from keys %sites instead of incrementing it. This is especially true if some $siteid is not unique.

$count=0;
$sth->bind_columns(\($siteid, $imageurl));
while($sth->fetch()) {
 $sites{$siteid}=$imageurl;
 $count++;
}
[download]

$sth->bind_columns(\($siteid, $imageurl));
while($sth->fetch()) {
    $sites{$siteid} = $imageurl;
}
$count = keys %sites;
[download]

$flag, $type, and $size don't have to be file scoped. If you reverse the logic of $flag and change its name to $error, the foreach block becomes easier to read.

my $error = 1;
if ( my($type, $size) = ( head($imageurl) )[0, 1] ) {
    if ($size < 25600) {
        if($type =~ /image\/(gif|jpeg)/) {
            print "OK ($size bytes, $type)\n";
            $error = 0;
        } else {
            print "Wrong file type ($type, must be image/gif or image/
+jpeg)\n";
        }
    } else {
        print "Image size exceeded ($size bytes, should be < 25600)\n"
+;
    }
} else {
    print "Error\n";
}

if ( $error ) {
    print FILE "update big_ass_table set imagedown=imagedown+1 where s
+iteid=$siteid and imageurl='$imageurl';\n";
} else {
    print FILE "update big_ass_table set imagedown=0 where siteid=$sit
+eid and imageurl='$imageurl';\n";
}
[download]