Google just released a white paper on image pattern recognition, yesterday, in Miami. Not really a new concept and already developped by a few companies, like Imense, in the UK, this however is a sign that Google is going beyond text tagging to retrieve images.
In brief, this technology uses the “knowledge” of Picasa, Google image and Panoramio to define, locate and “recognize” a landmark. It then clusters all similar images. In the long run, this could tell you what you are looking at ( if you are that kind of touris)
The technology has a success rate of 80% and is not ready for prime time. However, it is on track to something we have been talking about here, over and over again. The slow disappearance of text tagging in favor of image recognition. The time is not so far when search engines will index automatically the content of an image making manual keywording obsolete. It will also deeply affect how people will search for images, using more visual search then just plain text search. While objects and scenes will be the first step, it will take much longer to recognize concepts and ideas.
More detail info on the Google blog