Google Labs this week launched a new product called Similar Images, which allows users to search for images using others images instead of words. Until now searching for images with Google meant using text entered into a search bar to describe things, the results of which are only as descriptive as the words entered into the query and the words found on the page where the image is indexed by Google.
Google Similar Images changes that. With this feature, after entering in what one is looking for, the results all have a link below them that will find other images like that one. It uses image recognition technology to read the image and match it to others.
Another example of this is TinEye, a search engine that lets you upload your own image and then it matches it to other images on the web including ones that have been highly distorted or edited.
An important thing about both examples is that search is still evolving and even with Google's dominance there are new innovations to come. It also shows image recognition can be used accurately on a large scale when combined with search.
Here's an example of Google Similar Images at work:
When searching for "paris" one will get results for everything from the Eiffel Tower to Paris Hilton. After clicking the "Similar Images" link underneath one of those image results it will refine the results to just images that match that one specifically. Google Images also has a feature that will then display those photos according to whatever color is chosen.