I don't think it's expensive compared to using more interpolators.
>derivative maps approach(camera space) - 415 fps
>derivative maps approach(object space) - 385 fps
>tangent space approach - 425 fps
Interesting numbers. Looks like they are fairly close.
>But, Morten, is this correct to compute perturbing normal in view space when you have any scale factor in world matrix?
To some extent it's a matter of definition. There is nothing wrong imo with defining the perturbation as taking place in view-space.
That being said you are correct that if you intend to scale the object then to get the same visual impact from the bumps you'd also have
to scale up the bump-scale accordingly by the same object scale when perturbing in view-space. This bump scale represents the magnitude/length of the imaginary displacement
applied to the original surface.
If the scale is non-uniform then ideally you'd have to perturb in object space and transform as you have been doing. However, you could also simply do some approximate solution
which takes either the max or the average of the 3 object scales and then applies that to the bump scale and use it in the view-space perturbation.
Using non-uniform object scales from object space to view space is fairly uncommon in games though. Most games don't even support scaling going from object to view space.
>Is there any rule or maybe hint how to control user-scale for dBs and dBt or designers specify it "by sight" ?
Like I said the bump scale (not counting the dimension part) is actually the magnitude/length of the displacement to the original surface.
See figure 1 in the thesis I referenced --> http://image.diku.dk/projects/media/morten.mikkelsen.08.pdf
Does everything make sense?
