Parallax mapping
Shifts texture coordinates per pixel by a height map and the viewing angle, so the surface appears to have real depth.
Normal mapping fools the light but not the eye — the surface is still flat when you look at it edge-on. Parallax mapping adds the motion: from a height map and the view direction it offsets the texture lookup so raised parts occlude what lies behind them, just like real parallax. It was introduced by Kaneko and others in 2001.
The costlier variant, parallax occlusion mapping (Tatarchuk, 2006), ray-marches through the height field and delivers both self-shadowing and crisp silhouettes within the surface. Cheaper than tessellating real geometry, but it falls apart at the surface's edges, where there is nothing to offset against — which is why it shows best on floors, walls and other surfaces seen at a grazing angle.