How height steps of parallax occlusion works?

Hello, i’m working around improve my displacement shaders and depth offsets using pixel depth, vertex normals, etc. I’ve been creating some new functions that works fine, but to improve this i need to replicate the effect of the "min and max steps of Parallax Occlusion Mapping. I look in the parallax function, but i don’t understand exactly how it works, and i don’t need to just copy, i need to understand how it works to improve it and to know exactly what i’m doing. Someone can guide me for a right direction? Thanks for you attention.

243958-parallax.jpg

the content examples project has several examples in it you could check out.