Narrator:       This is Science Today. With a single photo or video clip as a template, software developed by engineers at the University of California, Santa Cruz can sift through thousands of images or videos online to pull out the ones that look like the template submitted. Peyman Milanfar says unlike existing methods of computer recognition of visual objects, their software does not need to go through an exorbitant amount of examples first.

Milanfar:        There's some interesting examples of this technology that one can think about; searching the Web for photos or videos with photos or videos, a real sort of visual Google is sort of the obvious application. Organizing pictures, whether on the Web or on your own disk drive into similar groups; finding similarity on social networks is another one and of course there's many obvious applications to security and surveillance. What we have is the fundamental core technology, which once it is married to technology having to do with databases and with large processing power is going to have a, hopefully, a global impact.

Narrator:       For Science Today, I'm Larissa Branin.