Surely those are just standard depth-maps as used every day in VFX for compositing? Each pixel has a z-depth as to how far it is from the camera, and using standard compositing software (Nuke) you can blur just a slice (depth-wise) of the image based on the z ranges?