A grad student and his professor have developed software that eliminates (some) people from Google Street View images. And that could be a boon to Google’s ongoing battles over Street View and privacy.
Arturo Flores, a computer science grad student at UC San Diego, and Serge Belongie, his professor, have published their findings (1.7mb PDF) in a research paper called “Removing pedestrians from Google Street View images.” The paper explains in detail how their software creates “ghost-free mosaics” by matching redundant pixels from different photos of the same location “to remove the pedestrian as if it had never been there.”
As the images above show, it’s not perfect. The photo on the right mysteriously shows a pair of sneaks with no body attached. Other images from the paper show what appear to be ghostly outlines walking down city streets. Professor Belongie realizes the software isn’t perfect. “An artist using photoshop could make a more aesthetically pleasing result,” he tells the International Business Times. The research paper says their technique may not work in some outdoor scenes, such as when there are too many people in the Street View image or when a pedestrian is moving in the same direction as the Street View car.
So, what does Google think of all this? A spokesperson told us this afternoon that they have no current plans to use this technology, but they may be interested in learning more. Professor Belongie tells us he has not been contacted by Google about the Street View project, but he doesn’t know if Flores has been contacted or not. (Update:: Flores also tells us he has not been contacted by Google.)
Street View has come under fire from privacy groups and various government agencies — particularly in Europe — for the past few years. Our Google Street View article library has plenty of background on that.