John Battelle points at a post from Thomas Hawk where he argues that image search would be improved through tagging but more important, through user ratings of pictures. Sure. That's an extension of the idea that any type of search results could be improved through making use of personalization user data. But the key is trusting those doing the ranking.
I've covered before that a small community like Flickr faces completely different issues related to spamming than the major players face. If user rankings drive a picture to the top of Google Images, then some people are going to look at how to manipulate those rankings. That's not to say you don't want to try it. You do, and it could definitely help. But it's not necessarily an out-of-the-box solution.
Also, while this might be the perfect picture for "bridesmaids" to Hawk and nearly 300 others who call it a favorite at Flickr, that doesn't mean it's actually the best picture on that topic. In fact, the other results he points at from Google, Yahoo and MSN are all great if you're a bride perhaps wanting various pictures of bridesmaids in dresses as you ponder your own choice.
In contrast, Hawk's perfect picture is terrible from that perspective. As always, relevancy is largely subjective. And what's awesome photography to a photographer may not be the perfect image to someone thinking from a different perspective.