1. The second one is a distance function distðÞ with scores,
which when they are nearer to 10, the closer the position p
is to the center position s within the filter area. The third
one describes a segmentation term segðÞ that scores high,
if the input disparity Dip at position p belongs with high
probability to the same image segment as the filtered
output value Dos at position s, and low if not. As usual, in
conventional bilateral filtering, this term is driven by the
difference between the color values Ip and Is at positions p
and s, respectively.