Modeling Anisotropic Surface Reflectance with Example-Based Microfacet Synthesis

Jiaping Wang 1 , Shuang Zhao 2 , Xin Tong 1 John Snyder 3 , Baining Guo 1

Microsoft Research Asia

1

Shanghai Jiao Tong University

2

Microsoft Research

3

Surface Reflectance

satin metal wood

Anisotropic Surface Reflectance

isotropic anisotropic

Our Goal

modeling spatially-varying anisotropic reflectan

ce

Surface Reflectance in CG

• 4D BRDF

ρ

(

o

,

i

) – B idirectional R eflectance D istribution F unction – how much light reflected wrt in/out directions

o i

Surface Reflectance in CG

• 4D BRDF

ρ

(

o

,

i

) – B idirectional R eflectance D istribution F unction – how much light reflected wrt in/out directions • 6D Spatially-Varying BRDF: SVBRDF

ρ

(

x

,

o

,

i

) – BRDF at each surface point

x

Related Work I

• parametric BRDF models – compact representation – easy acquisition and fitting – lack realistic details parametric model [Ward 92] ground truth

Related Work II

• tabulated SVBRDF – realistic – large data set – difficult to capture • lengthy process • expensive hardware • image registration light dome [Gu et al 06]

Related Work II

• tabulated SVBRDF – realistic – large data set – difficult to capture • lengthy process • expensive hardware • image registration light dome [Gu et al 06]

Microfacet BRDF Model

• surface modeled by tiny mirror facets [Cook & Torrance 82]

Microfacet BRDF Model

• surface modeled by tiny mirror facets [Cook & Torrance 82] normal distribution shadow term fresnel term

Microfacet BRDF Model

• based on Normal Distribution Function (NDF) – NDF

D

is 2D function of the half-way vector

h

– dominates surface appearance

Challenge: Partial Domains

• samples from a single viewing direction

i

– cover only a sub-region

h

 Ω of NDF – How to obtain the full NDF?

partial region partial NDF complete NDF

Solution: Exploit Spatial Redundancy

• find surface points with

similar but differently rotated

NDFs material sample partial NDF at each surface point

Example-Based Microfacet Synthesis

partial NDF to complete + other surface points + rotated partial NDFs = completed NDF

Comparison ground truth isotropic Ward model our model anisotropic Ward model

Overall Pipeline

• BRDF Slice Capture • Partial NDF Recovery • Microfacet Synthesis

Overall Pipeline

• BRDF Slice Capture • Partial NDF Recovery • Microfacet Synthesis

Device Setup

• Camera-LED system, based on [Gardner et al 03] Camera

Capturing Process

Overall Pipeline

• BRDF Slice Capture • Partial NDF Recovery • Microfacet Synthesis

NDF Recovery

• invert the microfacet BRDF model Measured BRDF ,  Unknown Unknown NDF Shadow Term  [Ashikhmin et al 00]

NDF Recovery (con’t)

• iterative approach [Ngan et al 05] – solve for NDF, then shadow term – works for complete 4D BRDF data 1.

2.

,  [Ngan et al 05]  [Ashikhmin et al 00]

Partial NDF Recovery

• biased result on incomplete BRDF data ground truth [Ngan et al. 05] NDF shadow term NDF shadow term

Partial NDF Recovery (con’t)

• minimize the bias – isotropically constrain shadow term in each iteration before constraint after constraint

Recovered Partial NDF

ground truth [Ngan et al. 05] our result

Overall Pipeline

• Capture BRDF slice • Partial NDF Recovery • Microfacet Synthesis

Microfacet Synthesis

partial NDF to complete Merged partial NDFs completed NDF

Microfacet Synthesis (con’t)

• straightforward implementation: For

N

NDFs at each surface point Match against

(N-1)

NDFs at other points In

M

rotation angles for alignment • number of rotations/comparisons:

N 2 *M ≈

5 × 10 11

(N ≈

640k

, M ≈

1k

)

Synthesis Acceleration

• • – complete representative NDFs only (1% of full set) For

N

NDFs in each surface point Match with

(N -1)

NDFs in other location In

M

rotation angles for alignment • times of spherical function rotation and comparison

N 2 * M ≈

5 × 10 11

N' 2 * M ≈

5 × 10 7

( N ≈ ( N' ≈

640k 6.4k

) )

Synthesis Acceleration

• • • • – complete representative NDFs only (1% of full set) For

N

NDFs in each surface point

(N -1)

NDFs in other location – In

M

rotation angles for alignment

N 2 * M ≈

5 × 10 11

N' 2 * M ≈

5 × 10 7

N'* log(N'* M) ≈

5 × 10 5

( N ≈

640k

( N' ≈

6.4k

) )

Performance Summary

• 5-10 hours for BRDF slice acquisition in HDR – 1 Hour for acquisition in LDR • 2-4 hours for image processing • 2-3 hours for partial NDF recovery • 2-4 hours for accelerated microfacet synthesis On a PC with Intel Core TM 2 Quad 2.13GHz CPU and 4GB memory

Model Validation

• full SVBRDF dataset [Lawrence et al. 06] – data from one view for modeling – data from other views for validation

Validation Result

Limitations

• • visual modeling, not physical accuracy single-bounce microfacet model – retro-reflection not handled • spatial redundancy of rotated NDFs – easy fix by rotating the sample

Rendering Result: Satin

Rendering Result: Wood

Rendering Result: Brushed Metal

Conclusions

• model surface reflectance via microfacet synthesis – general and compact representation – high resolution (spatial & angular), realistic result – easier acquisition:  single-view capture  cheap device  shorter capturing time

Future Work

• performance optimization – capturing and data processing • extension to non-flat objects • extension to multiple light bounce

Acknowledgements

• Le Ma for electronics of the LED array • Qiang Dai for capturing device setup • Steve Lin, Dong Xu for valuable discussions • Paul Debevec for HDR imagery • Anonymous reviewers for their helpful suggestions and comments

Thank you!