I was thrilled to read this evening in the New York Times that Google Inc. (NASDAQ: GOOG) introduced last week in Beijing a prototype of a precision image search engine. This prototype is intended to do for digital images on the Web what Google's original PageRank software did for searches of Web pages.

If it works, all I can say is that I am glad someone has figured out how to do this. A few years ago it occurred to me that it would a great security tool to be able to take a picture of someone and pop the image into Google's image search to try to match the image with a known person. This does not strike me as a particularly original idea since I feel as though I have seen the idea on countless movies featuring international intrigue.

In Beijing, the two Google scientists presented a paper describing VisualRank, an algorithm for blending image-recognition software methods with techniques for weighting and ranking images that look most similar. Most commercial image recognition software uses the text associated with the image for the search process. But VisualRank uses parts of the images.

This idea is not ready for prime time because it has only been applied to a fairly small sample of images due to the immense computing costs. Still, it brings closer to the present the day when image search could actually be useful. And that's something I'd like to see.

Peter Cohan is President of Peter S. Cohan & Associates. He also teaches management at Babson College and edits The Cohan Letter. He has no financial interest in Google securities.