Sorry for the confusion about "which type of parallax mapping I mean": To be fair, I'm not sure myself which is which: I know that parallax is the effect that, alongside bump mapping which creates directional shadows on the details of a texture, stretches and shrinks parts of a texture based on camera position / angle in order to create the illusion of depth. I imagine there are multiple ways of achieving that, and personally I'm not picky about which is used... the most realistic is obviously the best, but I assume performance also plays an important role.
No problem and sorry if i came across has a douch, btw normal mapping can also technically be considered bump mapping, both just simulate "bumps" on a 2D surface, bump mapping using a B&W texture was the first tech invented to simulate "bumps" (another very old tech that i don't seam to remember the name now, also made that but made everything look like shiny glass from any angle), then normal mapping was invented, visually a bump map and a normal map do the exact same thing, the diference, is that the B&W bump map holds extra info that was not really needed to simulate simple bumps, it has the ability to hold complex height info about the surface in all three components, x y z (that is why it is still used today for Tessellation for example), making the math more heavy then it needed to be, a normal map works just by using the normal's of the surface, simplifying the math (there the term normal mapping) and only on the x y components and making z (height) always be 1, visually this made no diference but changed everything on the performance front, so yes, performance does play a major role in the choosing of what technique to implement, specially those that make heavy use of ray tracing, like Self Shadowing Parallax Occlusion Mapping, Relief Mapping, Steep Parallax Mapping, etc, etc, so imo referring to the techniques by their names prevent confusion and headaches in the long run.
Also IMO both Parallax Mapping and Parallax Occlusion Mapping could be implemented, used perhaps has different LOD stages or different game quality stages, they can be coded to work with the same bump map, but one works on older PC's and is faster, but looks worse (at grazing angles), the other looks better but is slower and only run on "modern" shader model 3.0 GPU's and beyond.
Isn't the blue channel only storing the z-component of the normal?
Yes but afaik is just storing a flat height of 1 and most of the time the z-component is not even needed to do correct normal mapping, there the blue channel not being even used in some games, so, from what i know, a simple normal map, do not stores enough info to make a good B&W bump map (also called height map), but some do try to do it with varied success, i just don't know exactly how they do it, would love to know tho.
Edited by HMart, 11 May 2017 - 10:31 PM.