Measurement, Inverse Rendering

COMS 6998-3, Lecture 4

Motivations

• True knowledge of surface properties

• Accurate models for graphics

• Augmented reality, scene editing

Photorealistic Rendering

Materials/Lighting(Texture Reflectance[BRDF] Lighting)

Realistic input models required

Arnold Renderer: Marcos Fajardo

Rendering Algorithm

80’s,90’s: Physically based

Geometry

70’s, 80’s: Splines90’s: Range Data

Flowchart

Photographs

Geometric model

InverseRenderingAlgorithm

Lighting

BRDF

Flowchart

Photographs

Geometric model

ForwardRenderingAlgorithm

Lighting

BRDF

Rendering

Flowchart

Photographs

Geometric model

ForwardRenderingAlgorithm

BRDF

Novel lighting

Rendering

Next 3 slides courtesy George Drettakis

Taxonomy 1( , , , , , ) ( , , , , , )in outx y t x y t

General function = 12D

Scattering function = 9D

Assume time doesn’t matter (no phosphorescence)

Assume wavelengths are equal (no fluorescence)

Single-wavelength Scattering function = 8D

Assume wavelength is discretized or integrated into RGB(This is a common assumption for computer graphics)

( , , , ) ( , , , )in outx y x y

Taxonomy 2

Single-wavelength Scattering function = 8D

( , , , ) ( , , , )in outx y x y

Bidirectional Texture Function (BTF)Spatially-varying BRDF (SVBRDF) = 6D

Ignore subsurface scattering (x,y) in = (x,y) out

Bidirectional Subsurface ScatteringDistribution Function (BSSRDF) = 6D

Ignore dependence on position

Light Fields, Surface LFs = 4D

Ignore direction of incident light

( , , , )outx y

Texture Maps = 2D

Assume Lambertian

( , )outx y

3D

Assume isotropy

BRDF = 4D

Ignore subsurface scattering

( , ) ( , )in out

Ignore dependenceon position

2D

Measure plane of incidence

0D

Low-parameter BRDF model

Outline

• Motivation

• Taxonomy of measurements

• BRDF measurement

• Highlights from recent work

• Next week: paper presentations

Definition of BRDF

Sourcesrc

src

Detectordet

det

E

L

dA

dA

src

det det

rf

dA Next several slides courtesy Szymon Rusinkiewicz

Measuring BRDFs

• A full BRDF is 4-dimensional

• Simpler measurements (0D/1D/2D/3D) often useful

• Start with simplest, and get more complex

Measuring Reflectance

0º/45ºDiffuse Measurement

45º/45ºSpecular Measurement

Integrating Spheres

• Sphere with diffuse material on inside

• Geometry ensures even illumination

• More accurate measure ofdiffuse reflectance

Gloss Measurements

• Standardized for applications such as paint manufacturing

• Example: “contrast gloss” is essentially ratio of specular to diffuse

• “Sheen” is specular measurement at 85°

Gloss Measuements

• “Haze” and “distinctness of image” are measurements of width of specular peak

BRDF Measurements

• Next step up in complexity: measure BRDF in plane of incidence (1- or 2-D)

Gonioreflectometers

• Three degrees of freedom spread among light source, detector, and/or sample

Gonioreflectometers

• Three degrees of freedom spread among light source, detector, and/or sample

Gonioreflectometers

• Can add fourth degree of freedom to measure anisotropic BRDFs

Image-Based BRDF Measurement

• Reduce acquisition time by obtaining larger (e.g. 2-D) slices of BRDF at once

• Idea: Camera can acquire 2D image

• Requires mapping of angles of light to camera pixels

Marschner’s Image-Based BRDF Measurement• For uniform BRDF, capture 2-D slice

corresponding to variations in normals

Marschner’s Image-BasedBRDF Measurement

• Any object with known geometry

BRDF Measurement is Hard!

Reflectance modeling (diff +specular texture)

Input Synthesized

Sato, Wheeler, Ikeuchi 97

Image-based measurement of skin

Marschner et al. 2000

Inverse Global Illumination

Yu et al. 99

From a single image

Boivin and Gagalowicz 01

Original Photo

Assignment (by tomorrow)

• Brief e-mail of proposed project, partners, plan of action (milestones)

• Iterate by e-mail or schedule appointments later in the week

• 1-2 page proposal due next Wed. including an intermediate milestone (Oct. 23)