Seems very similar to parallax occlusion mapping, where you essentially use height and surface normal maps on a single plane mesh with very few triangles to simulate a very complex surface. This is accomplished by ray tracing each pixel that the plane is occupying to determine where the view ray would intersect with the height map how it would reflect based on the surface normal map. It's a very convincing effect unless you look at the surface at a glancing angle and realize that the geometry is 100% flat.
This might be a dumb or obvious comparison, but this is somewhat similar to how Wolfenstein 3D made its breakthrough to "3D" in a lot of ways. It wasn't 3D so much as a clever 2D hack, or "2.5D".
https://www.youtube.com/watch?v=6PpWqUqeqeQ