I don’t usually like to re-post things from other sites, but this one is too pretty to pass up. A group of researchers from the University of Toronto and INRIA in France have developed a nonlinear video interface tool that allows you to manipulate video in the context of object motion rather than through the standard, static time dimension. It looks really cool, and seems like a very natural way to scan through film. Here is a demo video:
Get the whole scoop here.
Update: I tried out their demo program with my own video. The first few seconds of the video are navigable, but it cuts off rather quickly. I think that this might be because the kitty crosses her own path so many times. Even so, it is pretty impressive that the program pulled that much data out of such poor quality video. Their demo videos worked much better, as would be expected :-). Note: I used CamStudio to grab the screen. Heres how it went: