CS 775: Advanced Computer Graphics

Lecture 19 : Ambient Occlusion, Photon Mapping and Point-based Global Illumination

CS775: Lecture 19

Ambient Occlusion

● Shadow due to ambient illumination

● How exposed a point is to ambient lighting?– Sample with inside-out with Monte-Carlo Techniques– Sample with outside-in with multiple renders

N−L

Ap=1π∫

ΩV p ,ωi( n̂⋅ωi)d ω

https://en.wikipedia.org/wiki/Ambient_occlusion

CS775: Lecture 19

Ambient Occlusion

http://blog.digitaltutors.com/understanding-ambient-occlusion/

CS775: Lecture 19 https://support.solidangle.com/display/AFCUG/Ambient+Occlusion

CS775: Lecture 19

Photon Mapping

● Photon Mapping Algorithm– Emit photons– Trace photons– Store photons

● Caustic photon map● Global photon map

– Gather photons

Lo( p ,ωo)=Le( p ,ωo)+∫Ω

f r ( p ,ωo ,ωi)L i ( p ,ωi)cosθi d ωi

CS775: Lecture 19

Photon Mapping– Emit photons– Trace photons

CS775: Lecture 19

Photon Mapping– Store photons

● Caustic photon map● Global photon map

CS775: Lecture 19

Photon Mapping● Reflected Radiance Estimate

L r( p ,ωo)=∫Ωf r ( p ,ωo ,ωi)L i( p ,ωi)cosθi dωi

L r( p ,ωo)=∫Ωf r ( p ,ωo ,ωi)

dΦ2( p ,ωi)cosθi d ωi dAi

cosθi dωi

L r( p ,ωo)=∫ f r ( p ,ωo ,ωi)dΦ2( p ,ωi)dAi

L r( p ,ωo)≈∑p=1

n

f r ( p ,ωo ,ωip)ΔΦ2( p ,ωip)

π r 2

CS775: Lecture 19

Photon Mapping– Gather Photons

CS775: Lecture 19

Photon Mapping

– 100000 photons, 50 photons in radiance estimatehttp://graphics.stanford.edu/courses/cs348b/lecture/biased

CS775: Lecture 19

Photon Mapping

– 500000 photons, 500 photons in radiance estimatehttp://graphics.stanford.edu/courses/cs348b/lecture/biased

CS775: Lecture 19

Photon Mapping

– 200000 global photonshttp://graphics.stanford.edu/courses/cs348b/lecture/biased

CS775: Lecture 19

Photon Mapping

https://graphics.stanford.edu/courses/cs348b-competition/cs348b-02/brasket_deleon/

CS775: Lecture 19

Point-based GI● Divide scene into surfels

● Each surfel is a small disk – point, radius, normal, outgoing radiance

● Record direct illumination on these surfels

● Splat these on the hemisphere around a point in the gather phase and add to compute GI

CS775: Lecture 19

Point-based GI

Point-Based Global Illumination for Movie Production. Per Christensen. Pixar Animation Studios. SIGGRAPH 2010 Course

CS775: Lecture 19

Point-based GI

Point-Based Global Illumination for Movie Production. Per Christensen. Pixar Animation Studios. SIGGRAPH 2010 Course

CS775: Lecture 19

Point-based GI

Point-Based Global Illumination for Movie Production. Per Christensen. Pixar Animation Studios. SIGGRAPH 2010 Course

CS775: Lecture 19

Point-based GI

Point-Based Global Illumination for Movie Production. Per Christensen. Pixar Animation Studios. SIGGRAPH 2010 Course

CS775: Lecture 19

Consistency and Bias

E [X ]=∫ ...

limN→∞E [X ]=∫ ...

● Unbiased – produces correct image on average.

● Consistent- image approaches correct image as some parameter is increased.

CS775: Lecture 19

Consistency and Bias● Biased estimate can be good.

– Can have lower variance than unbiased estimate, and may look better.

● Readhttp://www.cs.columbia.edu/~keenan/Projects/Other/BiasInRendering.pdf

Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16Slide 17Slide 18Slide 19Slide 20