Making a flat picture look like it has real depth, using a second picture as a map of how high each part stands.
The naive way is to shift the picture sideways by an amount proportional to height, which gives a mild sense of relief but nothing more. This does better. It follows the line of sight into the surface in thirty-two small steps, lowering an imaginary height as it goes, and asks at each step whether the surface has come up to meet it yet. The first step where the ground overtakes the ray is where you are actually looking.
That is what lets tall things properly hide what is behind them, instead of just sliding past. The final answer is refined between the last two steps so the hidden edge lands cleanly rather than in visible bands.
pivot picks which height stays put. Everything above it moves one way and everything below moves the other.