A little bit of a tangent, but my brother does a lot of bottlecap projects, like this:

https://fbcdn-sphotos-a-a.akamaihd.net/hphotos-ak-xpa1/v/t1....

...and this:

https://scontent-b.xx.fbcdn.net/hphotos-xfp1/v/t1.0-9/106227...

I did a very similar project, although I just used the average pixel color and had a randomized list to do a greedy algorithm.

As I was building the concept out, I made a version that output to Excel as opposed to a bottle cap rendering. The results were pretty fun...

http://huksu.com/pixelator/excel/

Sometimes I dump random images into peoples' Google docs.

Nice! I made something similar to teach myself Qt. It took an image and turned it into an offset grid of circles. You could adjust the area of the image assigned to each cap to see how many caps you need to get something recognizable. Then it would spit out a listing of how many caps you need for each color.

Just a heads up: the number of caps needed to make a recognizable david tennant in front of the tardis is very high. Table size becomes an issue. :)

Cool table, but side comment on the site. So there is a ton of unused pixels on either side of the 'source code' sections which are forced into bottom sliders to show inside their undersized boxes. Presumably the syntaxhighlighter js stuff could be configured to make bigger boxes? Really hard to read the code the way it is.

Very cool project. I love people, data, and beverages -- this project touches on all three!