Okay, last brainstorming for the day: Collaborative 日本語 flickr annotations and/or tags (whatever's searchable and allows for pointing to specific objects in pictures with numerous elements) to ease gathering of concrete vocabulary images for people who want to add them to their vocabulary cards and find Google Images a bit abstract and scattershot (and prone to the occasional disgusting images)?

Also, it's worth mentioning that only VERY CLEAR PICTURES should be used in a picture dictionary. Concrete nouns such as animal names work very well.