There was this idea floating around on some virtual reality forums some years back about an "adblock for real life" that would work in a very similar way with a head-mounted display. In terms of pattern recognition it is not at all difficult. Detection and tracking of a set of images is pretty much a solved problem in computer vision (doing it reliably in realtime is still somewhat of a black art). I like how this is framed as an art project, and it's a very cool proof of concept, though it's an issue that it doesn't black out logos until after they have been seen, and the blackout box blinking on and off as the detector passes its threshold only draws attention to those areas. A practical implementation could use inpainting algorithms (also built into OpenCV, and even provided in one of the samples) to fill the logoed areas with background-like texture, so that it's not an obvious "censored" black patch but a smooth transition.