8

Click here to load reader

University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell Photon Mapping and Irradiance Caching

Embed Size (px)

DESCRIPTION

University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell Bidirectional Paths flux (photons) radiance shadow ray

Citation preview

Page 1: University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell Photon Mapping and Irradiance Caching

University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell

Photon Mapping and Irradiance Caching

Page 2: University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell Photon Mapping and Irradiance Caching

University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell

Bidirectional Paths

x2

x1€

x0

y1

y2

y0

Light path

Eye path

Shadow ray

Page 3: University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell Photon Mapping and Irradiance Caching

University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell

Bidirectional Paths

flux (photons)

radiance

L(x i → x i−1) = Φ(y j ) f r(y j−1 → y j → x i)G(y j , x i) f r (y j → x i → x i−1)

shadow ray

y j€

y j−1

x i€

x i−1

G(y j , x i) =cos(θ j )cos(θ i)

x i − y j

2

cos(θ j ) = (y j → x i) •r n j

cos(θ i) = (y j → x i) •r n i

Page 4: University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell Photon Mapping and Irradiance Caching

University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell

Weighting the paths

Lp = wi, jLi, jj= 0

N j

∑i= 0

N i

Pixel estimate is weighted sum of path radiances

where

wi,N−ii= 0

N

∑ =1 for N = 0,1,2,3,K

Choice of weights is critical for low variancesee Veach for details

Page 5: University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell Photon Mapping and Irradiance Caching

University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell

Advantages and disadvantages

Bidirectional path tracingGood for small sources of strong indirect illumination (caustics)

Path tracingGood for outdoor scenes with only a small part of the model visible

Both have trouble sampling mirror images of causticsBoth need lots of samples to avoid objectionable high frequency noise

Page 6: University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell Photon Mapping and Irradiance Caching

University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell

Photon mapping

Cache light paths for efficiency?Decouple geometry and illumination

Don’t use illumination maps where number of samples needed is dependent on primitive size and scene complexity

Get rid of high-frequency noise efficientlyDensity estimation

However, density estimation produces biased (but consistent) samples

Page 7: University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell Photon Mapping and Irradiance Caching

University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell

Two pass approach

Pass one: photon tracingBuild the photon map data structure

Pass two: renderingTrace eye paths using photon map to produce radiance estimates efficiently

Page 8: University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell Photon Mapping and Irradiance Caching

University of Texas at Austin CS395T - Advanced Image Synthesis Spring 2007 Don Fussell