Google and Image Recognition

June 29, 2009

Not content with sophisticated image compression, Google continues to press forward in image recognition. Face recognition surfaced about a year ago. You can get some background about that home-grown technology in “Identifying Images Using Face Recognition”, US2008/0130960, filed in December 2006. The company has  long history of interest in non text objects. If you are not familiar with Larry Page’s invention “Method for Searching Media” US2004/0122811 was filed in 2003.

app of face recogniton

Source: Neven Technologies, 2006

The catalyst for the missing link between auto identified and processed images and assigning meaningful tags to images such as “animal” or “automobile” arrived via Google’s purchase of Neven Vision (originally I think the company used the “Eyematic” name. The switch seems to have taken place in 2003 or 2004.)

At that time, All Business described the company in this way:

Neven Vision purchased Eyematic’s assets in July 2003. Dr. Hartmut Neven, one of the world’s leading machine vision experts, led the technical team that created the original Eyematic system. Dr. Neven is also developing groundbreaking “next generation” face and object recognition technologies at USC’s Information Sciences Institute (ISI).

Google snagged with the acquisition the Eyematic patent documents. These make interesting reading, and I direct your attention to “Face Recognition from Video Images”, US6301370, which seems to be part of the Neven technology suite. The US patent document is – ah, somewhat disjointed.

Mixing Picasa, home grown technology, and the image recognition technology from Neven, Google had the ingredients for tackling a tough problem in content processing; namely, answering the question, “What’s that a picture of?”

Google provided some information in June 2009. A summary of Google’s image initiative appeared in Silicon.com, which published “Google Gets a New Vision When It Comes to Pictures”. (Silicon.com points to CNet.com which originally ran the story.) Tom Krazit reported:

Google thinks it has made a breakthrough in “computer vision”. Imagine stumbling upon a picture of a beautiful landscape filled with ancient ruins, one you didn’t recognize at first glance while searching for holiday destinations online. Google has developed a way to let a person provide Google with the URL for that image and search a database of more than 40 million geotagged photos to match that image to verified landmarks, giving you a destination for that next trip. The project is still very much in the research stage, said Jay Yagnik, Google’s head of computer vision research.

For me the key point in the Silicon.com story was that Google used its “big data” approach to making headway in image recognition. When matched to technology evolving from the FERET program, Google can disrupt a potentially lucrative sector for some big government integration firms.  The idea is that with lots of data, Google’s “smart software” can figure out what an image is about. Tapping Google’s clustering technology, Google’s Picasa image collection has been processed engineers to assign meaningful semantic tags to digital objects that don’t contain text.

Google purchased Neven Vision in 2006, the same year it snagged Transformics, another pivotal Google purchase. Google said in 2006 in “A Better Way to Organize Photos”:

Neven Vision comes to Google with deep technology and expertise around automatically extracting information from a photo. It could be as simple as detecting whether or not a photo contains a person, or, one day, as complex as recognizing people, places, and objects. This technology just may make it a lot easier for you to organize and find the photos you care about. We don’t have any specific features to show off today, but we’re looking forward to having more to share with you soon.

The soon translated into a two or three year digestion and integration process. Face recognition became available in 2008. With the comment reported by Silicon.com and Cnet.com, Google will be dribbling out more image magic in the the months ahead. The go slow and surround and seep strategies are alive and well at Google in my opinion.

The key point is that Google continues to blend internal and acquired technology, then allowing Google engineers years to refine, adapt, and experiment. Based on my research, 2006 continues to be a hallmark year for Google technology. Acquisitions since 2006 have been interesting but continue to fill gaps, not provide leapfrog capabilities.

As Google deploys image recognition and other technology on its platform, I pay attention to the clumps of technology Lego blocks that Google possesses. For example, in law enforcement and intelligence, Google wizards can explore combinations of Postini (email), Neven (image recognition, mobile visual search), Keyhole (maps, gps), and other bits and pieces. With modest effort, Google could roll out an integrated law enforcement and intelligence system. Business Week nosed around this topic, but the publication left the story hanging. For example, a follow up with Neven technology hooked into a weapon systems suggests interesting lethal and sub lethal uses of the technology. Think about virtual surveillance zones. Think about processing images and video for face datasets using Gabor Wavelet ‘face template’ or a similar technology. Established players in this specialized niches may want to pay closer attention to the Google.

Stephen Arnold, June 29, 2009

Comments

One Response to “Google and Image Recognition”

  1. randome dude thirthy on December 22nd, 2009 8:37 am

    i like this software beacause its really cool, aljhf;lands,fmvnasdlkfjpoqriepoiqepwaoripodfkpa;djkfal;kdjfpankaces pankakes

    …..you tolded me to speaking my minds,,,,, dawoodadinga

  • Archives

  • Recent Posts

  • Meta