Google already indexes the text buried inside your email messages. So it's no surprise that the search engine cum world power is fiddling with the idea of indexing the text buried inside your photos and videos.
Last summer, the Mountain View, California outfit filed a patent application with the World Intellectual Property Organization entitled "Recognizing Text in Images," and yesterday, this clever little proposal hit the web, describing a system that automatically identifies characters that turn up in everyday digital images nabbed by everyday digital cameras and camcorders.
"Text in a digital image derived, for example, from an urban scene (e.g., a city street scene) often provides information about the displayed scene or location," the application explains. "A typical street scene includes, for example, text as part of street signs, building names, address numbers, and window signs."
Naturally, the Googlies aim to index such text and make it searchable. "The method includes receiving an input of one or more image search terms and identifying keywords from the received one or more image search terms," is the rather unfortunate wording. "The method also includes searching a collection of keywords including keywords extracted from image text, retrieving an image associated with extracted image text corresponding to one or more of the image search terms, and presenting the image."
This sort of thing would be particularly, um, useful when tacked on to the company's latest attack on world privacy, Google Street View, which is all about collecting photos of "urban scenes" - not to mention the suburban variety. If it can pull text from Street View images, Google can make it even easier for someone to invade your privacy - or find a really bad hamburger.
"In one implementation, the extracted text results from text recognition of images derived from street scenes is indexed and associated with a mapping application," Google says. "A user of the mapping application can search for a location, for example, by business name, address, store hours, or other keywords. In addition to mapping the location for the user, the mapping application can retrieve images matching the user's search.
"For example, a user enters a search for a McDonald's in a particular city or near a particular address. The mapping application generates a map to the McDonald's as well as presents an image of the McDonald's. The McDonald's image is retrieved using the indexed text from the image identifying the McDonald's and location information associated with the image, which identifies the location of the particular McDonald's in the image."
And of course, Google is toying with the idea of using this system to display ads: "In one implementation, advertisements are presented along with the presented image. For example, an advertisement can be presented for the business identified in the image. Alternatively, one or more advertisements can be presented for alternative businesses. Additionally, the advertisement can be for one or more products associated with the business in the presented image, user search terms, or according to other criteria."
As Information Week notes, one of the Googlies behind the patent, Luc Vincent, refers to himself as "leader of several large geo-related projects, including Street View". Meanwhile, Vincent's patent-seeking partner, former Google intern Adrian Ulges, has helped develop a system that "autonomously learns to tag videos with high-level semantic concepts by watching videos from online portals like http://youtube.com." So your videos may be in danger as well.
Not that this patent application breaks entirely new ground. The folks at Riya already offer text recognition as part of their own new-age image search platform.
After launching a tool that did text and facial recognition on your private photo collection, Riya has now launched a kind of shopping engine that lets you locate products using a purely visual search. And it would seem that Google is considering a similar idea.
"In one implementation, a store (e.g., a grocery store or hardware store) is indexed," Google's application continues. "Images of items within the store are captured, for example, using a small motorized vehicle or robot...Text is extracted from the product images. In particular, extracted text can be filtered using a product name database in order to focus character recognition results on product names."
The difference, of course, is that we have no worries that little Riya will take over the universe. ®