I can imagine this being a substantial improvement to just bilinear filtering but how are the depths of continuous d determined? In three dimensions is there some kind of algorithm that studies the depth of a certain object and decides which level it should be at, and then everything else is interpolation?