bidirectional photon mapping jiří vorba charles university in prague faculty of mathematics and...

Post on 01-Apr-2015

224 Views

Category:

Documents

2 Downloads

Preview:

Click to see full reader

TRANSCRIPT

1

Bidirectional Photon Mapping

Jiří VorbaCharles University in Prague

Faculty of Mathematics and Physics

2

Motivation

Goal: Robust Global IlluminationPhoton Mapping good at:

CausticsReflections of causticsSmall light sources

Photon Mapping bad at:Glossy scenes

© H.W.Jensen

© Wojciech Jarosz

3

Our contribution

Photon Mapping

4

Outline

Review of Photon Mapping (Jensen et al.)Main ideaAlgorithm overviewResultsApplication to progressive methodsConclusion

5

Photon Mapping review

First passPhoton shootingOverall lighting approximation

© Kavita Bala

6

Photon Mapping review

Second passLight path connection heuristic

Primary rays

BRDF lobe

PMquery

Secondary rays

7

PM issues

1. Correlated queries

8

PM issues

2. Radiance estimate variance

9

Main idea

PMquery

PMquery

PMquery

10

1st strategy

PMquery

11

2nd strategy

PMquery

12

3rd strategy

PMquery

13

Strategy combination

Multiple importance sampling (Veach 97)

321

11

PPP

Pw

321

22

PPP

Pw

321

33

PPP

Pw

14

Algorithm overview

First passthe same as in PM

Second passPath-TracingRadiance estimate at every path vertexlight path contribution multiplied by

MIS weight

15

Results

Photon Mapping Our Bidirectional Photon Mapping

16

Application to progressive methods (not a part of the CESCG article)

Overcoming the memory limitsSolving finer detailsConsistent method

Our implementationBased on Stochastic Progressive PM

17

Progressive resultsOur progressive bidirectional method Our bidirectional method

(not progressive)

18

Results

Photon Mapping Progressive Bidirectional Photon Mapping

19

Reference - PT(320x320, 4 days on 6 cores)

PBDPM (512x512, 1 day 3h on 6 cores)

Results

20

Limitation

Bias introduced by an inappropriate MIS weights “gap” between a photon hit and a camera ray end point

PMquery

PMquery

Our assumption

Reality

21

Conclusion

ContributionNew more robust photon mapping based

method Future works

“Area light source” strategyCompensate the bias introduced by

inappropriate MIS weights

22

The end

Thank you!Questions?

Reference - PT(320x320, 4 days on 6 cores)

PBDPM (512x512, 1 day 3h on 6 cores)

top related