SIFT algorithm moves seamlessly from idea to iPhone to Interpol

If you take photos on an iPhone, you may very well have used The University of British Columbia’s most widely adopted invention.

It’s called Scale-Invariant Feature Transform, or SIFT for short. An algorithm developed in 1999 by UBC Computer Science Professor David Lowe, it has so far been licensed by more than 20 companies around the world and incorporated into dozens of different products.

SIFT is the invisible magic that matches shared elements in different images, helping amateur photographers create seamless panoramic photos; the visually impaired to navigate their environments; and retailers to reduce theft. Recently an Icelandic company, Videntifier, working in partnership with Interpol, licensed the algorithm to develop technology that will help combat the sexual abuse of children.

SIFT matches shared features from various images. When Lowe first published his SIFT findings, neither he nor UBC had any idea just how powerful this idea would prove to be. But the seminal research paper quickly gained star status and has since been cited over 50,000 times.

That attention alerted academics, industry and other inventors to the diverse possibilities for SIFT.

Since then, UBC’s University-Industry Liaison Office (UILO) has busily licensed SIFT technology to a wide variety of companies.

“It’s gratifying to see my work out there improving people’s daily lives in ways I had not anticipated,” says Lowe.

In 2003, Lowe and then-UBC PhD student Matthew Brown took advantage of the university’s dynamic Department of Computer Science research environment to explore their own novel application for SIFT.

The result was an invention they named AutoStitch—a panorama photo software system that automatically recognizes and matches images that are similar, even if they are unordered, have different exposures or come from different cameras. The software then “stitches” the images together to create seamless panoramas up to 360 degrees.

Half of the images stitched together
Stitched and blended
A 360-degree panorama created using AutoStitch to match, stitch and blend the images automatically.

Again with the help of UILO, AutoStitch was licensed and is now incorporated into dozens of applications. Lowe set up a UBC spin-off company, Cloudburst Research to create the Autostitch iPhone app that has been installed on more than half a million mobile phones and tablets around the globe.

And Lowe isn’t done yet. He sees even more applications for his original algorithm and dreams of an amped-up application that would literally put the world at travellers’ fingertips—or at least on their phones.

“Imagine yourself in the middle of a city and by taking just one photo with your phone, you can learn exactly where you are—essentially a computer vision approach more accurate than GPS. I think it’s within reach,” he says.

In the future, Lowe can envisage SIFT being used in self-driving car camera systems that will monitor surroundings and tell the car to put on the brakes when something, or someone, is in the way.

There’s certainly no stopping the forward momentum of SIFT.

Read more

 

The development of SIFT and AutoStitch was supported by funding from multiple sources including NSERC and its Idea to Innovation program.

 

Related Stories

Innovation + Impact