different mapping techniques

[Legend]

when looking at game pics, what to I look for do tell what kind of mapping is/isn't used? simple visual evidence.

 

[Xmas]

when looking at game pics, what to I look for do tell what kind of mapping is/isn't used? simple visual evidence.

- pre-baked lighting: can look good from a distance, but won't change when lights or viewer moves, so the lighting can look like coming from the wrong direction when this happens. Looks flat from an angle, no self-occlusion.

- emboss bumpmapping: looks acceptable when viewed flat-on, but gets worse quickly at a steeper angle. Only useful for larger bumps, but only small "depth". Looks poor and flat when viewed from an angle, no self-occlusion. Hardly ever used in games, irrelevant today.

- normal mapping aka dot3 bumpmapping: the "standard" bumpmapping. Looks quite good from a wide angle range, but still flat when viewed from an acute angle, no self-occlusion. Perceived depth greater than with emboss BM, but still not much. Self-shadowing is possible. Lighting changes when the light source moves (diffuse BM), or when either the light source or the viewer moves (specular BM). Specular BM is subject to shimmering artifacts, unfortunately.

- environment mapped bump mapping (EMBM): for highly reflective (or refractive) surfaces (water, glass, polished metal) with bumps, where the bumps "correctly" distort the reflection. Looks still flat from an angle, no self-occlusion.
The term EMBM can either mean a specific (fixed-function) implementation or simply all methods that try to simulate reflective bumps.

- parallax mapping: finally, proper self-occlusion which allows real, convincing depth.
Some variants allow a correct, bumped silhouette, but that is expensive and the silhouette cannot be antialiased with multisampling.

- displacement mapping: looks correct from any angle because it's "real" geometry. But usually done with relatively low resolution, so you could see the silhouette consisting of straight line segments. Silhouette gets antialiased with multisampling.


There are several variants of the above-mentioned groups of techniques, but the groups themselves are usually not hard to discern visually.

 

[Legend]

thanks alot

 

[Squeak]

... read that article.

CONCLUSION

We have shown that is possible to do DOT3 normal mapping very similar in quality to PC-style normal mapping ... w/out any additional effect on performance /normal mapping with ordinary shading/.

We have shown that it is possible to do DOT3 normal mapping very similar in
quality to PC-style normal mapping. Two methods have been presented: A
cheap two-pass solution, without per-pixel normalization, and a more expensive
four-pass alternative with per-pixel normalization. Both take advantage
of the hardware available on the PS2, more specifically the GS and the VU1.
This paper has also described how it is possible to integrate the normal
mapping with ordinary shading of non-normal mapped objects without any
additional effect on performance.

You could use the flexible geometryhardware to progressively subdivide the polygon(s) ~ perpendicular to the light vector, to achive a much better light-distribution than regular gouraud allows.

Also, I think the two-pass solution could be boiled down to one-pass, if only 128 light intensities where needed. This would be fine for anything but appearance-preserving on large objects.