The Google Visual Search research was revealed on a CNBC "Inside the Mind of Google" TV story that aired on Dec. 3, writes Clint Boulton in our sister publication, eWEEK. In a demonstration for CNBC interviewer Maria Bartiromo, Google Product Manager Hartmut Neven took a picture of Santa Monica Pier, then showed the technology positively identifying the landmark and delivering relevant search results.

In the CNBC clip, Neven reportedly mentions applications such as taking a picture of a painting in a museum to bring up additional information, or photographing a French menu to get a quickie translation.

In 2006, Google acquired Neven's startup Neven Vision, which made facial and image recognition biometric software. The software was initially targeted at image recognition for Google's Picasa photo-sharing application, but has since broadened into other applications. Code-named Google Goggles, the technology is based on a Visual Mobile Search (VMS) service that returns search results based upon image recognition.

According to eWEEK, Google Goggles "didn't pass muster" when Google tested it with a focus group in August. But the company is pushing ahead, "working out the bugs and building out the immense database required to propel the technology."

Image recognition beyond faces is still a major challenge, both at the algorithm level and in designing large enough databases to identify a wide range of objects at every angle. Because cameraphone search results can be linked to GPS coordinates, it's possible that Google Visual Search results could be refined, for example, helping tourists to identify major landmarks.

The service could also provide additional information about surrounding attractions, as well as location-based marketing offers. Conceivably Google Visual Search could be combined with "augmented reality" apps that display retrieved geo-locational information overlaying a photograph, among other possibilities.

Advertising may be Google's main goal for the technology, writes Boulton, suggesting the company might pair image-based search with digital ads, possibly from Google's AdMob acquisition. Neven has filed some patents for broader applications, including one for image-based contextual advertisement method and branded barcodes, writes Boulton.

In the CNBC clip, Neven reportedly said, "Visual advertising content may be displayed on a digital billboard or large television screen. A user may take a picture of the billboard and the displayed advertisement to get additional information about the advertised product, enter a contest, etc."

Such a targeted use of the technology would likely be easier to implement than a broad-based visual search service.

Other groups have also been working on the intersection of camera-phones and image databases. For example, at MIT's CSAIL, the Vision Interface Group has done work on a somewhat similar project called IDeixis. The IDeixis technology is being considered for applications ranging from tourism to real estate (photographing a house for sale to bring up more information) to interactive marketing (snap a shot of a passing car to bring up photos and specs).

Meanwhile, as Boulton points out, visual search recognition apps, such as ViPR from Evolution Robotics, are being offered for the Apple iPhone.

Availability

Google Visual Search will eventually come to Android phones, although Neven did not mention a timetable, says eWEEK. Nor did he say whether the technology would be made available to other smartphone platforms.

The eWEEK story on the technology may be found here. A CNBC page touting its "Inside the Mind of Google" series may be found here.

• Winning Android apps announced
• Android app offers remote medical diagnostics
• Android Challenge winners push location awareness
• Ten Android Challenge winners featured
• Linux robot car targets autonomous navigation
• Android phone sports 8.1-megapixel camera
• HD surveillance camera design runs homespun Linux