Dirty Glass: Rendering Contamination on Transparent Surfaces - PowerPoint PPT Presentation

1 / 61
About This Presentation
Title:

Dirty Glass: Rendering Contamination on Transparent Surfaces

Description:

Dirty Glass: Rendering Contamination on Transparent Surfaces – PowerPoint PPT presentation

Number of Views:107
Avg rating:3.0/5.0
Slides: 62
Provided by: cave52
Category:

less

Transcript and Presenter's Notes

Title: Dirty Glass: Rendering Contamination on Transparent Surfaces


1
Eurographics Symposium on Rendering 2007
Dirty Glass Rendering Contamination on
Transparent Surfaces
Jinwei Gu Ravi Ramamoorthi
Peter Belhumeur Shree Nayar
Columbia University
2
Transparency in CG Renderings
Hakura and Synder, 01
Whitted, 80
Jensen, 01
Zongker et.al., 99
Chris Wyman, 05
Narasimhan et.al., 06
Jensen, 00
3
Transparent Objects in Real World
Courtesy of many photographers from flickr.com
4
Transparent Objects in Real World
  • Effects
  • Contrast reversal

Dirty Window
5
Transparent Objects in Real World
  • Effects
  • Contrast reversal
  • Angular dependency

Stains on Wine Glass
6
Transparent Objects in Real World
  • Effects
  • Contrast reversal
  • Angular dependency
  • Lens glare and blur

Image Taken with a Dirty Lens
7
Transparent Objects in Real World
  • Effects
  • Contrast reversal
  • Angular dependency
  • Lens glare and blur
  • Monitor effects


8
Transparent Objects in Real World
  • Effects
  • Contrast reversal
  • Angular dependency
  • Lens glare and blur
  • Monitor effects
  • Simply alpha-blending with a texture will NOT
    work, because contaminants will both scatter and
    attenuate light.



9
Related Work
  • Surface Aging Weathering
  • Patina, weathered stone, flow, etc.
  • Surface cracks, scratches, etc.

J.Dorsey et al., 00
T. Wong et al., 97
S. Merillou et al., 01
Weathering on opaque surfaces.
  • Light Scattering in Layered Materials
  • Simulation of dusty surfaces
  • Reflection of layered materials
  • Rendering of composite paint

J.Blinn, 82
P. Harahan and W. Kruger, 93
S. Ershov et al., 01
Additional scattering components are required for
contaminated transparent surfaces.
10
Our Work
  • Rendering contamination on transparent surfaces
  • Modeling Analytic BRDF/BTDF Model based on
    Single Scattering
  • Measurement Contaminants Patterns Scattering
    Parameter
  • Rendering 3D Scenes and 2D Photographs

11
Our Work
  • Rendering contamination on transparent surfaces
  • Modeling Analytic BRDF/BTDF Model based on
    Single Scattering
  • Measurement Contaminants Patterns Scattering
    Parameter
  • Rendering 3D Scenes and 2D Photographs

12
Clean Transparent Surface
Li
Air, n1
Transparent Medium, n3
Snells law Fresnel equations
13
Contaminated Transparent Surface
Air, n1
Li
Contaminant, n2

Transparent Medium, n3
14
The BRDF/BTDF Model
  • The contaminant layer is optically very thin.
  • e.g., dust, lipids/ fingerprint, dirt/salt, etc.
  • Single scattering dominates.
  • Aggregate as a (local) BRDF/BTDF model.

15
Single Scattering Events Reflection
16
Single Scattering Events Transmission
17
The BRDF/BTDF Model
  • BRDF
  • BTDF

Optical thickness of the contaminant layer
Heyney-Greenstein phase function
(Scattering) albedo of the contaminant
Refraction index of the contaminant
18
The BRDF/BTDF Model
19
The BRDF/BTDF Model
20
The BRDF/BTDF Model
Change the incident angle
21
Verification
Proposed Model
Monte Carlo
4 samples/pixel,7 seconds
2048 samples/pixel,1 hour
RMS relative difference 2.2
22
Comparison with Related Work
Reflectance Lr
  • Collimated beam
  • Albedo 0.5
  • Optical thickness 0.2
  • Phase function g0.9
  • Refraction index 1.0

Monte Carlo
Our Model
HK93/Blinn82
Outgoing Angle
23
Our Work
  • Rendering contamination on transparent surfaces
  • Modeling Analytic BRDF/BTDF Model based on
    Single Scattering
  • Measurement Contaminants Patterns Scattering
    Parameter
  • Rendering 3D Scenes and 2D Photographs

24
Measurement (1)
  • Optical thickness
  • Setup
  • Shadow images due to attenuation

Lambertian board
Contaminants
Thin glass
Camera
Projector
25
Measurement (1)
  • Optical thickness
  • Setup
  • Shadow images due to attenuation

26
Some Acquired Textures
Household Dust
Powder
Water Stripes
Soil
Fingerprints 1
Fingerprints 2
Smudges
Detergent
Salt Water Deposit
Soap Water
Oil Stain
Salt Deposit
27
Measurement (2)
  • Scattering parameter
  • Setup

Thin glass
Camera
28
Measurement (2)
  • Scattering parameter
  • Setup

29
Results
Clean
Fingerprints
Soap Deposit
Sand
30
Results
Clean
Fingerprints
Soap Deposit
Sand
g0.99
g0.94
g0.86
g0.78
31
Synthesis Tool
32
Our Work
  • Rendering contamination on transparent surfaces
  • Modeling Analytic BRDF/BTDF Model based on
    Single Scattering
  • Measurement Contaminants Patterns Scattering
    Parameter
  • Rendering 3D Scenes and 2D Photographs

33
Clean Window of a Room
Outside the room
Inside the room
34
Dirty Window View inside the Room
Contrast Reversal
35
Dirty Window Viewed outside the Room
36
Dirty Window in 24 hours
37
Contamination on Monitor Screen
Monitor ON
Monitor OFF
Monitor Effects
Angular Dependency
38
Stains on Cognac Glass
Contrast Reversal
39
Stains on Cognac Glass
Vertical Shadow
40
Stains on Cognac Glass
Angular Dependency
41
Stains on Cognac Glass
Without Scattering
With Scattering
42
Rendering on 2D Photographs
Image
Lens
Contaminant layer
Planar Light Source
(2D Photograph)
43
Seawater Sprayed on Glass
Photograph
Contaminant Texture
http//www.flickr.com
44
Seawater Sprayed on Glass
Photograph
http//www.flickr.com
Composition
45
Dirty Camera Lens
46
Making a Window Dirty
Homography camera position
Contaminant texture
Photograph
F. Durand and J. Dorsey, 02
Synthesized light probe
47
Making a Window Dirty
Photograph
F. Durand and J. Dorsey, 02
Composition
48
Conclusions
  • Analytic BRDF/BTDF model based on single
    scattering.
  • Simple measurement for various kinds of
    contaminants.
  • Rendering applications on 3D scenes and 2D
    photographs.

Source code measured data available
at http//www.cs.columbia.edu/CAVE/projects/dirty
_glass/
49
Future Work
  • Contaminated translucent materials
  • e.g., lipids on skin
  • Cleaning dirty glass in photographs

50
Acknowledgement
  • Srinivasa Narasimhan, Stephen Marschner, Craig
    Donner, Bo Sun, Kevin Egan, Aner Ben-Artzi, Ryan
    Overbeck, Li Zhang, Kshitiz Garg for insightful
    discussions and help.
  • Anonymous reviewers for detailed comments and
    suggestions.

51
Thank you!
  • Comments Questions?

52
(No Transcript)
53
Q A
54
Measurement (1)
  • Why only consider attenuation?
  • Most contaminant are mainly forward scattering
    (ggt0.7).
  • If too much blur, we use another setup below and
    scale the result accordingly.

Lambertian board
Contaminants
Thin glass
Camera
Projector
55
Comparison with Related Works
  • Collimated beam.
  • Albedo 0.5
  • Optical thickness 0.2
  • Phase function g0.9
  • Refraction index 1.0

56
Thin Glass Slab with Contamination
  • BRDF/BTDF model by similar derivation

Li
Air, n1
Contaminant, n2
Transparent Slab, n3
Air, n1
57
Dusty Glass Sphere
58
Dust on Cognac Glass
59
Comparison with Related Works
Blinn 82 Hanrahan Kruger 93 Our Model
Derived From Geometric Inference Transport Theory Transport Theory
Fresnel No Yes Yes
Lr(1) No Yes Yes
Lr(2) No No Yes
Lr(3) Yes Yes Yes
Lr(4) No No Yes
Lt(1) Yes Yes Yes
Lt(2) Yes Yes Yes
60
Derivation of the BSDF Model (1)
  • Assume the contaminant layer has no interface
    between the air (e.g. dust)

61
Derivation of the BSDF Model (2)
  • Assume the contaminant layer has no interface
    between the air (e.g. dust)

n1
n2
Write a Comment
User Comments (0)
About PowerShow.com