It's probably the getimagesize function - it is going and fetching every image on the page so it can determine the size. Maybe you can write something with curl to get the header only for Content-size (though, actually, this might be what getimagesize does).

At any rate, back in the day I wrote a few spiders and it's kind of slow to do, with internet speeds better than ever it's still a fetch for each element. And I wasn't even concerned with images.