BURLINGAME, Calif.--Google has begun testing face-blurring technology for its Street View service, responding to privacy concerns from the search giant's all-seeing digital camera eye.
The technology uses a computer algorithm to scour Google's image database for faces, then blurs them, said John Hanke, director of Google Earth and Google Maps, in an interview at the Where 2.0 conference here.
Google has begun testing the technology in Manhattan, the company announced on its LatLong blog. Ultimately, though, Hanke expects it to be used more broadly.
Dealing with privacy--both legal requirements and social norms--is hard but necessary, Hanke said.
"It's a legitimate issue," he said. He likened the issues some have with Street View to the ones that took place when Google introduced aerial views to Google Maps. It took time for the public, regulators, and Google to get comfortable with the feature, but, "It needs that debate. We see that and try to let it play out."
New jurisdictions, new rules
Street View poses other privacy issues besides just faces. Some people aren't eager to have their houses on display, for example. But much of the hubbub seems to have waned since Google launched Street View in May 2007, and indeed other companies such as Blue Dasher are working on similar technology.
Street View presents a view of dozens of United States cities from a driver's perspective (unless a plastic bag is stuck over the Street View camera). It appears Google has begun collecting imagery in Europe as well, along with detailed 3D maps, including Milan, Rome, and Paris.
A Pittsburg couple sued Google for allegedly photographing images on a private drive in April, but it's legal to take photos from public streets in the United States. However, standards vary.
"A just balance needs to be found between what can be publicized, in deference to the principles of freedom of expression and of information, and what has to be safeguarded from excessive public curiosity, so as to avoid infringing the individual's right to privacy and right to his or her picture," the French embassy observes.
Years of research
The face-blurring technology took a year to develop and is based on prior research that took several more years, Hanke said.
Face detection, which humans perform effortlessly with help from some dedicated neurons in the visual cortex, is a decades-old computer science problem. It's finally arriving in basic form in real-world applications, though, including digital cameras that use it to track and properly expose subjects or take a picture only when subjects are smiling.
There are some potential complications for Google Street View, though. False positives that blur billboards or works of art with faces could degrade Street View a bit, but missing some faces that are visible could pose privacy problems.
Google thinks its technology has struck the right technology balance in general.
"It does a good job of figuring that out. It uses a variety of technologies to filter," Hanke said, though it's "not perfect."
Many times computer algorithms struggle to recognize faces that aren't straightforward views. But that problem isn't as bad for Google: the faces that are obscured by hair, telephone poles, or oblique views are likely the ones identifiable already.
Have you found any examples of faces the algorithm missed or that it should have caught? Share the links or other thoughts in the comments section below.