Recovering BRDF Models for Architectural Scenes

1
Recovering BRDF Models for Architectural Scenes
SIGGRAPH 2000 Course on Image-Based Surface
Details

• Yizhou Yu
• Computer Science Division
• University of California at Berkeley

2
Image-based Rendering versus Traditional
Graphics ( circa 1997 )
Improved photorealism - Static scene
configuration - Fixed lighting condition
3
Image-based Modeling and Rendering

• Vary lighting
• Recover reflectance properties for multiple
  objects in a mutual illumination environment

500am
600am
700am
1000am
4
The Problem

• Forward Problem Global Illumination
• Couple lighting and reflectance to generate
  images
• Backward Problem Inverse Global Illumination
• Factorize images into lighting and reflectance

Illumination
Reflected Light
Reflectance
5
Global Illumination
Reflectance Properties
Light Transport
Images
Geometry
Light Sources
6
Inverse Global Illumination
Reflectance Properties
Images
Geometry
Light Sources
7
Input Images
Every surface should be covered by at least one
photograph A specular highlight should be
captured for every specular surface
8
Camera Radiance Response Curve

• Pixel brightness value is a nonlinear function of
  radiance.
• Debevec MalikSiggraph97 gives a method to
  recover this nonlinear mapping.

Intensity
Saturation
Radiance
9
In Detail ...
10
Recovered Geometry and Camera Pose
11
Light Sources
Spherical light sources are easier to model Light
source intensity can be calibrated from dynamic
range images
12
Synthesized Images
Original Lighting
Novel Lighting
13
A Comparison
Hand-crafted
Recovered
14
Outline

• Diffuse surfaces under mutual illumination
• Non-diffuse surfaces under direct illumination
• Non-diffuse surfaces under mutual illumination

15
Lambertian Surfaces under Mutual Illumination

• Bi, Bj, Ei measured
• Form-factor Fij known
• Solve for diffuse albedo

16
Parametric BRDF Model Ward 92
N
H
Isotropic Kernel
( 3 parameters)
Anisotropic Kernel
( 5 parameters)
17
Non-diffuse Surfaces underDirect Illumination
N
P2
H
P1
P2
P1
18
Non-diffuse Surfaces under Mutual Illumination

• Problem LPiAj is not known.
  ( unlike diffuse case, where LPiAj
  LCkAj )
• Solution iterative estimation

Source
Aj
LPiAj
LCkAj
Pi
Target
LCvPi
Ck
Cv
19
Estimation of Specular Difference S

• Estimate specular component of by
  Monte Carlo ray-tracing using current guess of
  reflectance parameters.
• Similarly for
• Difference gives S

Aj
LPiAj
LPiAj
Pi
LCkAj
Ck
LCkAj
LCvPi
Cv
20
Recovering Diffuse Albedo Maps

• Specular properties assumed uniform across each
  surface, but diffuse albedo allowed to vary.
• Subtract specular
  component
• Recover pointwise
  diffuse albedo

21
Results

• A simulated cubical room

22
Results for the Simulated Case
Diffuse Albedo
Specular Roughness
23
Results

• A real conference room

24
Real vs. Synthetic for Original Lighting
Real
Synthetic
25
Diffuse Albedo Maps of Identical Posters in
Different Positions
Poster A
Poster B
Poster C
26
Inverting Color Bleed
Input Photograph
Output Albedo Map
27
Real vs. Synthetic for Novel Lighting
Real
Synthetic
28
Modeling Outdoor Illumination

• The sun
• Diameter 31.8 seen from the earth.
• The sky
• A hemispherical area light source.
• The surrounding environment
• May contribute more light than the sky on shaded
  side.

29
A Recovered Sky Radiance Model
30
Coarse-grain Environment Radiance Maps

• Partition the lower hemisphere
  into small regions
• Project pixels into regions and
  obtain the average radiance

31
Comparison with Real Photographs
Synthetic
Real
32
Inverse Global Illumination

• Detect specular highlights on the surfaces.
• Choose sample points inside and around
  highlights.
• Build links between sample points and facets in
  the environment
• Assign to each facet one photograph and one
  average radiance value
• Assign zero to Delta_S at each link.
• For iter 1 to n
• For each link, use its Delta_S to update its
  radiance value.
• For each surface having highlights, optimize its
  BRDF parameters.
• For each link, estimate its Delta_S with the new
  BRDF parameters.
• End