[GAMES 202] Lecture 9

Screen Space Directional Occlusion

Paper

Use direct lighting information, from rendering of screen space. (Instead of RSM, reflective shadow maps)

3D Space Method: At shading point $p$, shoot a random ray: -> hit obstacle: indirect illuminated -> no hit: no indirect illumination

The inverted version of SSAO! (shade color when no ambient lightingin SSAO) Insight: SSAO assumes indirect lighting comes from distant surfaces, while SSDO assume it from nearby surfaces.

Steps: For each shading point $p$, sample points in the upper sphere near $p$. For a sampled point $q$, visibility $V(p\leftrightarrow q)$ is approximated with $V_{\text{cam}}(q)$.

Cons: Short ranged GI Limited to screen space Visibility is approxinated

Screen Space Reflection

SIGGRAPH Course

Method: Screen space ray marching using normals and depth buffer!

Implementation details: At each step $p_i$ along a ray, check $p$‘s depth and depth buffer, until finds an intersection Step size: trade off between quality and computational costs Glossy surfaces? trace a lobe according to BRDF(use importance sampling)

Adaptive step size: Depth mipmap using mean pooling (instead of average) Using the consecutive property of a node’s children. (similar to BVH or a binary search tree)

Cons: Limited to screen space (hidden geometry, edge cutting, etc.)