You’ve probably noticed Google’s rolled out an ingenious way to create bokeh via depth maps in pictures without any additional hardware (it’s available now in the Play Store). HTC’s One M8 has a dedicated depth sensor, and Samsung has a software based pseudo-blur effect, but Google’s solution produces results on par, or better than them, and most importantly, it’s available everywhere now.
I am thoroughly impressed, and here’s a technical description of the algorithms here, involving Markov Random Fields. Naturally, the next question is, can this technology be used retro-actively, to create 3D models based on old videos, to some degree? Already, alternate applications of the depth maps are available, such as this web-based parallax picture viewer. Very cool.