Google Blogoscoped

Thursday, May 05, 2005

Image Comparison Engines

“CBIR” is the handy acronym for Content based image retrieval. Basically, it means instead of entering text to find images, you provide or click on an existing image to find related ones. This ought to be much more intuitive for certain tasks.

A CBIR research project of Penn State University has now been applied to an aviation images database, Slashdot reports. Click on “Show me photos”, and then click on “View similar photos” to get an idea of how well this works. This is not necessary related to actual image recognition (analyzing a picture to find out it contains, say, an elephant), but can be implemented using much more brute force pixel-by-pixel image comparison with some added mirror and scaling fuzzyness.

If Google or Yahoo would add this feature to their image search, they’d have an immediate killer advantage to other image searches – and lawyers hunting down copyright infringements would have a new powerful tool at their hands.

Or... did Google already have this in place? One reader by the nick of Chefmonkey at Slashdot comments:

“Google actually did take this technology and try it. The first version of their image search had a “find similar” link next to every image. These tended to work okay at first (they weren’t great, but you usually got enough photos back that you could visually scan them and find something of interest that was related to the original image). After a few months, for some reason, the “find similar” links started returning increasingly nonsensical results. After it degenerated to the point of near uselessness, they took the “find similar” link away from the image search results. I expected it to turn up again once they got the kinks worked out, but apparently they just decided to stop working on it.”

Image Comparison Engines by Philipp Lenssen | Comments









Pixel Group

Brain Games




Text Link Ads
Movable Type Weblog
Tutorials, samples and more.