14 DAY TRIAL // Google’s got bigger problems with Street View at the moment, the Wi-Fi debacle may have quieted down but it’s not gone, but for a long time the biggest concern was the fact that people’s faces were sometimes visible in the images that landed online. Google does a fairly good job at blurring people’s faces and car number plates to protect privacy, but it may eventually eliminate people altogether from the photos.
Cnet dug up a research paper by a computer science graduate which describes a technology which ‘deletes’ people from Street View imagery. The technology works by getting information from the several photos of the point which Street View usually takes as the cars pass by.
A prototype algorithm has been demoed and it actually does a decent job. It’s not perfect by any means, there are tell-tale artifacts left in most of the images which have had people removed from them. The algorithm analyzes several views of the same region. It finds the common spots, then removes the human figures and fills in the space with processed imagery from the other photos. But this approach has limitations, it relies on the data already available, if it can’t find a different view for the region it tries to fill in, there’s not much it can do.
Interestingly, removing people from images is not as impressive today as it might have been just a few months ago and we’ve got Adobe to thank for that. The latest version of the Adobe Photoshop editor comes with the ‘content-aware’ technology which enables users to fill in a portion of a photo with content as determined by a very clever algorithm. Results vary, but it can be quite amazing at times.
Now, if Google really wanted people gone from Street View photos, it could blend the two methods and come up with a way of cleaning up the photos that doesn’t leave a trace. The question is, would removing people from those images actually make sense.