NeoMedia Technologies, Inc (NEOM)

[dlethe01]

PP:When Is A Camera Phone A Web Browser?

Sunday, January 15, 2006
Soon your camera phone will be used as a browser by recognizing images as hyperlinks.

A couple of the physical world connection players (Mobot and Neven Vision) are highlighted in this story (sub req'd).

Camera phone shots used for web searches

If your phone had image-recognition software, you could take a picture of the place and use it to instantly pull up a list of reviews. And you could do the same for monuments and tourist attractions, CDs, logos, billboard advertisements, almost anything you want information on.

Using an image is a convenient way to bring up information, especially as phone keypads are not really designed for typing.

See how valuable physical world hyperlinks will be?
http://theponderingprimate.blogspot.com/2006/01/when-is-camera-phone-web-browser.html
posted by Vangorilla @ 5:02 PM
******************************************************

Camera phone shots used for web searches
07 January 2006
Celeste Biever
http://www.newscientist.com/channel/info-tech/mg18925335.700
http://www.scanbuy.com/website/news_newscientist.htm
Why type a web search into your mobile phone when you can simply show it what you want to find?

YOU'RE walking down the street looking for somewhere to have dinner. You stop outside a nice-looking restaurant, but there aren't many people in it. Is that because it's too early in the evening, or because the food and service are terrible?

If your phone had image-recognition software, you could take a picture of the place and use it to instantly pull up a list of reviews. And you could do the same for monuments and tourist attractions, CDs, logos, billboard advertisements, almost anything you want information on.

Using an image is a convenient way to bring up information, especially as phone keypads are not really designed for typing. Using a picture is also a more succinct way to sum up what it is you are looking for - if you know the exact CD or book you would like to buy, for example.

Advertisers are already beginning to use visual cellphone searching in promotional campaigns. These encourage consumers to photograph a product, logo or billboard advertisement using their phone and receive a gift in return. And a further raft of consumer-oriented applications, such as accessing online stores, holiday guides and movie times, is on the way.

"The camera phone is the first networked camera," says Anita Wilhelm, who has founded a company called Caterpillar Mobile in San Francisco, with the aim of incorporating camera phones into multi-player games. "People are just starting to realise the potential."

One of the first ideas was to link barcodes to web addresses. Software in the phone would recognise a picture of the barcode and take the user to the relevant site. For example, New York-based Scanbuy introduced a product called ScanZoom in February 2004, which takes shoppers to the relevant Amazon.com page when they photograph the barcode on the back of a book or CD, allowing them to buy it just by taking the photograph.

But this idea is limited to products that happen to have bar codes, says Lauren Bigelow at the mobile visual search company Mobot, based in Lexington, Massachusetts.

So Mobot and Neven Vision of Santa Monica, California, are using object-recognition software. Advertisers pay to enter products, logos, movie posters or magazine ads into a database. When people photograph them and send the image to Mobot or Neven as a picture message, they receive information such as a web address or text message in return.

Neven's first campaign in October, called CokeFridge, encouraged readers of German teen magazines to snap photos of a Coca-Cola logo in return for a free soccer computer game. Mobot has enabled US camera-phone users to buy ring tones and music through snapping a photo of a CD cover. These promotions are an advertiser's dream, because the success of an ad can be measured by counting the number of times people send in a photo of it.

Phones will soon be able to recognise text within photographs, too. The company 23half in Mountain View, California, has released a prototype system called nThrum, which translates the image of text into actual text, and then feeds it into a search engine. This allows you to carry out a much broader search, as it does not rely on a central database of stored information. "We are trying to provide useful information to end users," says 23half's CEO, Kumar Gopalakrishnan. In future the system could also be used to translate text, such as signs in foreign countries, he says.

Later this year Bandai Networks, the largest cellphone applications provider in Japan, will enable users to buy products by taking photos of them, using 3D object-recognition software provided by Evolution Robotics in Pasadena, California.

Mobot and Neven are planning to create further consumer-oriented services, including being taken automatically to your favourite sports team's website when you take a photo of their logo, or a travel guide in response to a photo of a monument in a city. Neven already has a prototype system that allows people to buy from an online catalogue by photographing a product, and it plans to add text recognition too, to improve the system's overall ability to recognise objects.

Object-recognition software divides the 2D image captured by the camera into a grid, and then assigns positions to thousands of "interest points" containing distinctive features such as textures or angles. These are used to create a unique signature for the image. To match two images, the software looks for similar interest points in similar positions.

This method is very good at matching identical objects, such as logos, movie posters, CD covers and monuments, and in the past has been used by Sony's robotic dog Aibo to identify its charging station. It is also used in a robotic camera made by Evolution called Lane Hawk, which sits at supermarket checkouts and monitors the shelf underneath trolleys to spot stolen goods.

But what would be really useful for mobile visual search would be to go beyond simple object matching to genuine object recognition. Then software fed an image of a chair, say, would not just identify it, but also assign it to a category. Right now this cannot be done. "A chair is not an object, it's a concept," says Munjal Shah, who has designed an automatic photo-archiving system called Riya that uses face-recognition software. "And concept recognition is extraordinarily difficult."

If it were possible, the uses for mobile search could significantly broaden. If you took a picture of a car, for example, the system could search for all local car dealers, rather than that specific make or brand.

Adding context information, such as the location at which the photograph was taken, would improve the search further. "Then you truly have a search engine in the palm of your hand," says Paolo Pirjanian, CEO of Evolution.

The returns could be huge. The number of people carrying camera phones worldwide hit 370 million in 2005 and looks set to more than double in 2006, according to market research firm InfoTrends/CAP Ventures in Weymouth, Massachusetts. "It's the most personal device. Next to underwear there is nothing else on our bodies for longer than this," says Roger Entner, a telecoms consultant at Ovum in Boston, Massachusetts.

Lookalike trouble

Celeste Biever

At the moment the object-recognition software used by Evolution Robotics and Neven can pick a match from a database of around 100,000 possibilities. This is peanuts compared with Google's 8 billion indexed pages.

But simply increasing the number of photos in the database wouldn't work, says Andrew Cory, CEO of Neven, because the search engine would begin to make mistakes. "When you make larger databases it is more likely that objects will look similar," says Cory. And the more alike each other the objects appear, the more difficult it is for the system to recognise them.

Sorting the objects into categories would narrow down the search before the object-recognition software kicks in. In this way Neven hopes to increase the size of the database the software can cope with by a factor of 10 every three months.

But even with this improvement Cory believes image search is unlikely to completely replace word searches on cellphones. "If you are looking for sports trivia it might be hard to find an image that will take you there," he says.