(Annette Moser-Wellman) -- Does your Web site allow users to upload their photos? Or do you have a large photo and video inventory? If you answered yes to either of these questions, read on. Recognition technology is rapidly evolving in ways that could create new opportunities for your media organization.

So, first off, when I hear 'recognition software,' I immediately think of Tom Cruise getting his retina scanned in Minority Report. But the technology to recognize people is quickly moving beyond security screening and launching into applications that offer fresh ideas for managing video and photos and even extracting value from them.

Eyealike is one such recognition software company, which seeks to take recognition technology to a new level. President Greg Huess told me about his firm's histogram technology, developed in partnership with the University of Washington. By grabbing four or five bits of information from a photo or video, Eyealike creates a "fingerprint" that allows a server to understand its content. Rather than relying on tagging (assigning a word to a piece of content to enable it to be categorized and searched), the histogram lets the computer identify what's in photo itself. Is this a face? Is this a face of a baby? Is there an animal in the photo? What kind? Is there an object such the Eiffel Tower in the photo?

Huess said, "We build out a 3D model or rendering of a face and then we look deeper – what's the length of the hair, the color and texture? What's the skin tone? What's the geometry of the face? Are there dimples and a square jaw? We put all that data in our backend algorithm and spit out the match results."

This kind of image recognition capability is in its infancy on the consumer front, with products like Riya using image recognition to enable consumers to sort pictures more easily. The major difference between Riya and what Eyealike is envisioning is the sheer volume of scanning that allows them to filter massive files – like video on YouTube. And Eyelike can recognize objects. They're also talking about linking it to advertising.

So now that you have an idea how the technology works, imagine the advertising possibilities. Any site or page with photos and videos, such as social networks or photography sites, could be "read" and used for targeted advertising. For example, if someone has uploaded a lot of baby photos, that user may well be interested in baby products. Or if someone has uploaded pictures of the Taj Mahal, perhaps this user is interested in travel ads.

Major social networks with hundreds of millions of photos have already spotted the potential of this technology, approaching Eyealike with ideas on how to target users by the content of images. Remember what retailing pioneer John Wanamaker once said, "Half the money I spend on advertising is wasted; the trouble is I don't know which half." With the growth of these recognition technologies on the Web, advertisers might soon know the answer to that question!

And for those with big libraries of video and photo, recognition software could help with the massive job of making these libraries searchable. For example, for those in the news business, many photo and video inventories are not tagged. What if you could call up images from the library without using tags? Do you want an image of the Golden Gate bridge or a corporate logo? Recognition software could allow you to scan libraries to find and call up the specific content desired, without going to all the time and expense of tagging all those images.

Eyealike hopes to revolutionize search with these recognition technology tools. You may be aware that Google has a massive program underway to get users to tag photos; recognition software could render that effort unnecessary. Huess shares, "In the next few years, search is going to be more and more image-based. For example, you're going to type in 'little black dress' and the engine will pull up every little black dress at every store in your area."

What is particularly interesting is that the same technology can be reverse-engineered to block content from being uploaded. Consider networks and movie studios. Huess says, "If you can imagine pulling down copyrighted material from a YouTube that has 170,000 uploads every single day, you have to be able to do it very quickly and, on top of that, you have to be able to do it accurately." With a claimed 95% accuracy rate, Huess' histogram technology may be able to provide protection for intellectual property that has plagued networks and movie studios since the dawn of Web 2.0.

Because it's possible to block images being uploaded in a more accurate way, social networks in China have expressed interest in using the technology to try to abide by the restrictions of the government.

"In China, the population under 30 doesn't watch TV at all. They are doing everything online. People are putting user-generated content up on these sites and social networks are running into government pressure to filter content from over 200 million users. I know that sounds absolutely horrible to us in the states, but technology like ours makes it possible to manage a bit of the chaos," Huess said.

It's been said that a picture is worth a thousand words. It's this power that the Web is just beginning to harness. It will be an exciting ride to see how the meta-data available in recognition software will shape that journey.


What do you think? Please share your thoughts, experiences and reactions by clicking on the comment button below.


Annette Moser-Wellman is President of Firemark, Inc., an innovation consultancy, and author of "Running While The Earth Shakes: Creating An Innovation Strategy To Win In The Digital Age," published by Media Management Center. She teaches in MMC's Advanced Executive Program and Digital Strategies for Media Executives seminars.

This TechScout article is part of a new series of Moser-Wellman interviews commissioned by Media Management Center to explore opportunities and insights at the intersection of technology and the news media. Click here to view others in the TechScout series.