Penn State researchers have announced an upgraded CAPTCHA approach, based on images.

First, you have to figure out where the overlapping images' borders are, so you can find the center of one of them and click near it. The images have been distorted and pixillated somewhat.
The fact that this is difficult speaks to how hard it is to track photo usage in layered composite images, which we see a lot of every day.


Having done that, you now must associate another image with a term that would logically apply to it. The example shown is a garlic bulb, in case you can't tell - so the term from the list that's most applicable is "vegetable".
Now, this bears a striking resemblance to keywording, which is a workflow challenge for all of us in the graphics industry. So if anyone cracks this (and frankly I hope they can't because I need a working CAPTCHA more than I need automated keywording) I have work for them...

Link

via Slashdot Post

Background on the project and team here