
Large, powerful companies making people 'disappear' is usually the stuff of Hollywood (though we take our lives in our hands every time we write a negative review), but if one computer science graduate has his way, Google might do just that.
Arturo Flores has developed an algorithm that scans images from Google's ubiquitous Street View service, identifies passing pedestrians and takes a digital paintbrush to them. It uses the patterns and colours surrounding the offending individual to mask their presence from the Goog's all-seeing eye.
Flores' concept explores the feasibility of removing pesky pedestrians, who may block potentially useful views of landmarks or locations. He published his findings in June and demonstrated the system at the IEEE Mobile Workshop on Mobile Vision. Despite claiming the results are largely 'ghost-free', in the images we've seen there's a palpable, creepy outline of the victim -- though that may just be us backmasking.
You'd think this system would fall flat in a crowded area, where there's insufficient information on what's surrounding the teeming masses of, say, old London town. Though, according to Flores, the algorithm is best suited for densely populated urban areas because pavements, for example, are usually a consistent colour. That doesn't quite sound right -- as his test shots show, most people photographed by the Google cars are at least partly in front of buildings.
Google has no plans to run this system through Street View, nor has it commented on Flores' finding in any way. If it does, the novelty of finding yourself ambling about town will be lost. But if you're one of the unfortunate folk who felt the intruding gaze of Street View cars at precisely the wrong moment, that might be a relief.