
3. Volumetric Fog and Lighting 237
Regular grid sampling Jittered grid sampling
Figure 3.13. Comparison of regular and jittered sampling patterns.
one layer of objects is stored (the depth buffer acts like a height field—we have
no information for objects behind it). Therefore, when reprojecting a dynamic
2D scene, occlusion artifacts are inevitable and there is a need to reconstruct
information for pixels that were not present in the previous frame (Figure 3.14).
In the case of volumetric reprojection, it is much easier, as we store informa-
tion for whole 3D viewing frustum in volumetric textures, as well as for the space
behind the shaded objects.