DigitalBeat

Google scientists announced a new technology today that might bring big improvements to how Google processes and understands images. For now, it’s limited to images of landmarks like the Golden Gate Bridge and the Eiffel Tower, but conceptually, at least, its promise is much broader.

Basically, a user can upload an image, then Google uses this new technology to analyze it and automatically identify if it’s a photo of a landmark, and if so which one. Google says it created a list of landmarks using tagged photos in Picasa, Panoramia, and online tour websites, then pulled the best images of each landmark from those sites plus Google Image Search, and finally developed “a highly efficient indexing system for fast image recognition.” For now, this isn’t a product, just a paper presented at the Computer Vision and Pattern Recognition conference in Miami (click here to download the PDF). But this could lead to automatic photo-tagging in Picasa, as well as more search results in Google Image Search.

Now, you might not be someone who cares about uploading pictures or finding photos of landscapes. If so, the more exciting question is whether this technology could be used in other areas. Right now, Google Image Search is based largely on how websites describe images in tags, leading to a lot of random results. But the technology described here might enable Google to group photos of people or things based on what’s actually shown in the image. (Startups such as Eyealike are also using image recognition technology for similar purposes.)

Technologies like this, as well as the “similar images” feature that’s being tested, as well as the VisualRank concept that Google presented in a paper on last year, could lead to a much more powerful version of Image Search, and one that Google could make more money from through shopping or e-commerce.

Next Story: SpinVox brings voice-to-text overseas with $100M Telefonica deal
Previous Story: Six Apart bloggers get VideoEgg’s interactive ads