the "ground truth" method

Hi, can anyone please tell me how the method of “ground truth” works when generating soft shadows? Thanks.


I’m guessing that you mean “ground plane” shadows. This simple technique involves a projection of geometry on to a plane, presumably the ground, followed by a blending operation.

For a light direction L of unit length, and a 
vertex V projected on to a plane with normal N 
and distance D, the projected vertex V' is given 
         D - <V|N>  
V' = V + ---------  L
For a point light with position L, the projected
vertex V' is given by
         D - <V|N>
V' = V + --------- (L - V)
         <L - V|N>

where <*|*> is a standard dot product.   

There are alternatives to this technique, such as stenciled shadow volumes and depth buffers, both facilitated in current hardware. The shadow volume method uses the same basic projection concept as the ground plane method above.

There are numerous soft shadow techniques. One method involves jittering a shadow volume eminating from the light source; another solution utilizes precomputed light maps for static scenes.

This is currently a huge research area, and there are several interesting ideas out there.

I hope this helps.