Here's the basic logic. You'll want to insert whatever extra logic -- but this won't stop anyone from downloading it, as once it's in their browser, they can do anything they want with it. (it can be used to check to see if someone's directly linking to an image, if the browser sends HTTP_REFERER)

#!/usr/bin/perl --

open (IMAGE, '<', '/path/to/image')
    or print "500 Server Error\n\n" and exit;
print "Content-type: image/gif\n\n", <IMAGE>;
close (IMAGE);
[download]

In reality, you want to send the correct MIME type for whatever image you're sending, and send additional headers (cache-control, size of the file, etc.)

The times that I've done image rotating scripts, I've saved a copy of the image, with full HTTP headers, so I could just dump out the file contents without needing to do any extra parsing to generate headers each time around.