[scallywag]

Thought this might be of some interest to people here:

http://www.nature.com/news/2004/041122/full/041122-3.html#top

and

http://www.cs.dartmouth.edu/~farid/publications/pnas04.html

and

http://science.slashdot.org/article.pl?sid=04/11/24/0258257&tid=126&tid=14

[Odds]

Sounds interesting, but I'm way too lazy to read all that; mind giving us a little summary?

[scallywag]

Hrum. Okay.

It's a mathematical way of analysing pictures based on the small-scale details in them.

This is a trivial example... but say you like to shade (in line drawings) diagonally from bottom left to top right. The value in any pixel will be much better correlated with the values in the pixels below-left than they will in the pixels below right,

They are talking about a way of "fingerprinting" this charactristic of your values structure... and a hundred other similar characteristics.

It's like a spectoral analysis of your value structure, not the big blocks that determine the piece but the choices you make all the way through. They are mapping you in the frequency domain - analysing how closely placed your brush strokes are rather than where they are placed.

Of course (like any scientific analysis) the method of analysis that they have chosen is to some extent subjective. Don't get frightened by the big words here. Basically what they have done is found a clever way of saying "shit, there's a lot of diagonal lines in this pic" (or horizontal lines, or dark blobs, or whatever). Then they produce a "fingerprint" based on this, then they compare this (in glorious unlabelled 3D graphs) to other "fingerprints" of other artworks. Do they match? Yes! Is this signifcant? Perhaps.