Scalable Virtual Ray Lights Rendering for Participating Mediathachisu/vrlcuts_slides.pdf · 2019....

Scalable Virtual Ray Lights Rendering for Participating Media

Nicolas Vibert Adrien Gruson Heine Stokholm Troels Mortensen Wojciech Jarosz Toshiya Hachisuka Derek Nowrouzezahrai

McGill University Luxion VIA University College Dartmouth College The University of Tokyo

MOTIVATION: Surface and Volume interaction2

MOTIVATION: Volume interaction only3

VOLUMETRIC RENDERING4

Sensor

Image Plane𝜔

𝐿𝑚(𝑥, 𝜔) = න0

𝑠

𝑇𝑟(𝑢)𝐿𝑖(𝑥𝑢,𝜔)𝑑𝑢

𝑥𝑢

s

VOLUMETRIC RENDERING: Rendering techniques

• Path tracing / Bidirectional path tracing

5

VOLUMETRIC RENDERING: Many techniques6


• Density estimation:• Volumetric Photon Mapping

• Photon Beam

• Photon Planes

• “Higher-order geometric primitives”

VOLUMETRIC RENDERING: Many techniques7


• Density estimation:• Volumetric Photon Mapping

• Photon Beam

• Photon Planes


• Many lights:• Virtual point lights

• Virtual spherical lights

• Virtual ray lights


MANY LIGHTS

• Many-light techniques have been introduced in “instant radiosity” [Keller et al. 1997]

• Indirect illumination as a sum of direct illumination of virtual lights

8

𝝎

MANY LIGHTS9

Virtual point light contribution

𝐿𝑚𝑉𝑃𝐿 =

)𝑉𝑃𝐿(𝑦, x𝑢)|𝑦 − 𝑥𝑢|

2𝑝(x𝑢

𝑦

𝑥𝑢

𝐿𝑚𝑉𝑃𝐿 =

)𝑉𝑃𝐿(𝑦, x𝑢)|𝑦 − 𝑥𝑢|

2𝑝(x𝑢¥

MANY LIGHTS10

∝1

𝑦 − 𝑥𝑢2

𝑥𝑢

𝑦

Virtual point light contribution

MANY LIGHTS

• VPL vs. short VRL

11

MANY LIGHTS: VRL

Virtual ray lights contribution

12

𝐿𝑚VRL = න

0

𝑠

න0

𝑡 )𝑉𝑅𝐿(𝑢, 𝑣

𝑤 𝑢, 𝑣 2d𝑣d𝑢

𝑢

𝑣𝑉𝑅𝐿

MANY LIGHTS: VRL

Virtual ray lights contribution

13

𝐿𝑚𝑉𝑅L =

)𝑉𝑅𝐿(𝑢, 𝑣

)𝑤 𝑢, 𝑣 2𝑝(𝑢, 𝑣

𝑝 𝑢, 𝑣 ∝ 𝑤 𝑢, 𝑣 −2

𝑢

𝑣𝑉𝑅𝐿

MANY LIGHTS: VRL vs. VPL14

VRL VPL

Equal rendering time

MANY LIGHTS15

Realistic rendering : • Unmanageable amount of virtual lights• Cost linear with lights

MANY LIGHTS16

Aim: • Sub-linear cost• Scalable methods

[Walter et al. 2005][Walter et al. 2006][Walter et al. 2012][Hasan et al. 2007][Ou et al. 2011][Bus et al. 2015]

Realistic rendering:• Unmanageable amount of virtual lights• Cost linear with lights

SCALABILITY17

SCALABILITY18

INDIVIDUAL LIGHTS

SCALABILITY19

CLUSTERS

INDIVIDUAL LIGHTS

SCALABILITY20

CLUSTERS

INDIVIDUAL LIGHTS

RELATED WORKS21

Previous works have already explored a combination of VRLs with scalable techniques:

• Adaptive light-slice for virtual ray light [Frederickx al. 2015]

• Adaptive matrix column sampling and completion for rendering participating media [Huo et al. 2016]

Our solution: Upper bound22

)𝑉𝑅𝐿(𝑢,𝑣

)𝑤 𝑢,𝑣 2𝑝(𝑢,𝑣< B

B=?

Our solution: Upper bound

Φ𝑓(𝑢, 𝑣): Constant within the VRL cluster

23

Φ𝑓(𝑢,𝑣)𝑇𝑟(𝑢)𝑇𝑟(𝑤 𝑢,𝑣 )




𝑇𝑟(𝑢): Impossible to control, assuming worst case => 1

24






𝑇𝑟(𝑤 𝑢, 𝑣 ): Based on the min distance

=> ℎ𝑚𝑖𝑛 ≤ 𝑤𝑘(𝑢, 𝑣)

=> 𝑇𝑟(𝑤 (𝑢, 𝑣)) < 𝑇𝑟(ℎ𝑚𝑖𝑛)

25

AABB

ℎ𝑚𝑖𝑛

𝑣

𝑢

𝑤 (𝑢, 𝑣)





𝑇𝑟(𝑤 𝑢, 𝑣 ): Based on the min distance

=> ℎ𝑚𝑖𝑛 ≤ 𝑤𝑘(𝑢, 𝑣)

=> 𝑇𝑟(𝑤 (𝑢, 𝑣)) < 𝑇𝑟(ℎ𝑚𝑖𝑛)




AABB

ℎ𝑚𝑖𝑛

𝑣

𝑢

𝑤 (𝑢, 𝑣)


𝑤𝑘 𝑢, 𝑣 2𝑝 𝑢, 𝑣=𝑤𝑘 𝑢, 𝑣 2𝑝(𝑢|𝑣)𝑝(𝑣)


Worst case when VRL and sensor ray are parallel.

)𝑝(𝑣 <1

𝐿𝑚𝑎𝑥, with 𝐿𝑚𝑎𝑥 the maximum length inside the cluster.

28



Worst case when VRL and sensor ray are parallel.

)𝑝(𝑣 <1

𝐿𝑚𝑎𝑥, with 𝐿𝑚𝑎𝑥 the maximum length inside the cluster.

29



AABB𝑣

𝑤 (𝑢, 𝑣)

Θ

𝑢

തℎ

ത𝑢

𝑠


AABB𝑣

𝑤 (𝑢, 𝑣)

Θ

𝑢

തℎ

ത𝑢

𝑠

𝑤 𝑢, 𝑣 2𝑝 𝑢 𝑣 = ഥℎ(ഥℎ2+ഥ𝑢2)

Θ(ഥℎ2+ഥ𝑢2)


AABB𝑣

𝑤 (𝑢, 𝑣)

Θ

𝑢

തℎ

ത𝑢



= ഥℎ

Θ

𝑠


AABB

ℎ𝑚𝑖𝑛



= ഥℎ

Θതℎ > ℎ𝑚𝑖𝑛

𝑠

Θ


AABB

ℎ𝑚𝑖𝑛



= ഥℎ


𝑠

Θ


AABB

ℎ𝑚𝑖𝑛



= ഥℎ


𝑠

Θ𝑚𝑎𝑥


AABB

ℎ𝑚𝑖𝑛



= ഥℎ

Θതℎ > ℎ𝑚𝑖𝑛Θ < Θ𝑚𝑎𝑥

𝑠

Θ𝑚𝑎𝑥


B = Φ𝑓 𝑢,𝑣 𝑇𝑟 hmin ΘmaxLmax

ℎ𝑚𝑖𝑛

Our solution: Light tree38

Agglomerative approach [Walter et al. 2008]:

- Not multithreaded

- Does not scale with high node overlapping 𝑂(𝑁2)

What we need:

- Fast/Parallelizable

- Agglomerative principal

Our solution: Light tree

Step 1: Partition the space with sorting

40


Step 2: Build local light tree that minimize the metric


Step 3: Build final tree with agglomerative process

Results

• Equal time comparison• VPL with LC

• VRL

• Two metrics• RMSE: sensitive to fireflies

• SMAPE: robust to fireflies

• Isotropic medium, only medium-medium interactions

43

Results44

Reference VPL LC (1M) 344 secs

RMSE: 9.01SMAPE: 3.25

VRL LC (100k)370 secs


VRL (10k)323 secs


Results45

VPL LC (1M) - 147 secsRMSE: 0.33

SMAPE: 2.44

VRL LC (100k) - 121 secsRMSE: 0.01SMAPE 1.88

VRL (10k) - 147 secsRMSE: 0.05SMAPE 9.11

0.0

> 0.05

SMAPE

Results46

Re

lati

ve s

pe

ed

-up

STARCAISE VEACH-LAMP VEACH-LAMP BATHROOM

VRL count VPL count

Summary

Contributions:

• New bound for VRL cluster• Efficient tree construction• X10 Speedup

47

Questions ?

48

Scalable Virtual Ray Lights Rendering for Participating Mediathachisu/vrlcuts_slides.pdf · 2019....

Documents

Transcript of Scalable Virtual Ray Lights Rendering for Participating Mediathachisu/vrlcuts_slides.pdf · 2019....