This section contained a number of different talks, only two of which I will comment on, due to lack of space and interest in the other two. The two that were interesting involved printing with metallic ink patterns and a very interesting new way to compare trees of data.
Reproducing Color Images with Embedded Metallic Patterns
If you're looking for cool output, this is the talk. The folks who were giving this presentation (from EPFL in Switzerland) are involved in, among other things, anit-counterfitting. It appears that the work came out of a need to model accurately the output colors and specular characteristics of metallic inks in order to be able to determine fill patterns in building new currency. The math was pretty detailed, but the effects were cool, in that they were able to substantially increase the accuracy of the transparent ink/paper model as well as the transparent ink/metallic ink/paper model from an error rate of 79% (that's 21% success) to 9% in coverage and reflection prediction. This has allowed them to create some truly stunning output with metallic watermark and highlighting features.
The one real visualization component of this talk, Tree Juxtaposer is designed to compare two trees of highly related data for differences. The difficulty comes when the trees have a large number of nodes, and they're working with some pretty big data sets (phylogenetics @ 300,000 nodes and IP connectivity maps at millions). They demonstrated a method and a java implementation, of their algorithm for locating and quickly exploring trees with differences using relatively low-resolution display devices (like CRTs). One of the more interesting aspects is that they did not reorder the trees, so the dual-column navigation technique will automatically zoom to the focussed node on one side and the corresponding node on the other side, based on content, not file position.
This would be particularly cool for looking at snapshots of IP connectivity graphs taken over successive days or using different methods.