If they are correctly determining something is rewritten, then 90% of the big media news and newspaper sites ought to be in the dumpster now, and the news/press release services would be at the top, but they are not. There are ONLY so many ways to say something. "John S., 99, died today of cancer. He's survived by... He was CEO of..., He invented the... etc." FACTS are FACTS! There is NO WAY to avoid repeating part of what someone else wrote if you are all trying to provide the same FACTS. Unless one of you is WRONG or just making stuff up - in which case... do you really want the "different" one to be on top? (Hmm, perhaps time for an experiment there ;-) If everyone is rushing to tell about the same "scoop", someone, perhaps even the first to know about it and the first to broadcast/print it, but having the slowest web team or the last to be crawled by G, is going to have to be last. G's tried various things to tell who was first, apparently without luck so far, so WHO should be penalized for "rewriting"? Perhaps both or all parties equally. Perhaps that IS what Panda is about, SPLIT the credit equally for repeated content among all sources? Otherwise, I'm still of the opinion that duplication issues in Panda are 95% restricted to same-site content duplication.
Mentioning images, I just noticed today (I'm probably a bit behind the curve) that on-site google adsense search is now putting small (90x90?) versions of photos from the page they return in the results when there is one and they think it applies to the term. It works about 50% in so far as the photo actually represents the keyword being searched for, especially when there are multiple photos and topics on the page (a daily news page for instance). So they're apparently taking more interest in photos for sure.