Last Time: Photometric Stereo: shape from lighting' Today, a little bit more about light, and images

1
Last Time Photometric Stereo shape from
lighting.Today, a little bit more about light,
and images.
2
(No Transcript)
3
Distant objects appear Bright !
4
(No Transcript)
5
De-hazed
Haze
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
Bad Weather
Mist
Haze
Rain
Fog
Images Courtesy Steve and Carol Sheldon
10
More Weather
Non-uniform Fog
Rain Drops and Rain Streaks
Snow Flakes and Snow Streaks
11
How often do we see Bad Weather?
Clear Sunny (77)
Bad Weather (23)
Manhattan, Every Hour, 12 Months
12
Natural illumination in Scattering Media
Narasimhan and Nayar, 99 - 03, Schechner et al,
01, 04
13
Glows of Light Sources
Mist
Fog
14
Active illumination in Scattering Media
Levoy et al., Narasimhan-Nayar, Kocak-Caimi,
Jaffe et al., Schechner et al., Negahdaripour et
al.
15
Floodlighting is Bad in Scattering Media
Remember Driving in Fog at Night?
16
Translucent Objects
Koenderink and van Doorn, 2001
Clouds
Milk
17
Rendering Moon
Jensen et al., 2001
18
Scattering in different fields
Art - 500-600 years Physics - 250
years Astrophysics/Astronomy - 80-100
years Atmospheric Optics - 80-100
years Medical Imaging - 30 years Remote
Sensing - 30 years Oceanic Engineering - 30
years Computer Graphics - 20 years Computer
Vision - 5-10 years
19
Light Transport in Clear Day
Point Source
Viewer
Surface Point
Near-Field Divergent Sources
20
Light Transport in Scattering Media
Point Source
Viewer
Surface Point
Clear Day
Foggy Day
Clear Day
Foggy Day
21
Complexity of Rendering Scattering Media
Objects
Virtual Viewpoint
Virtual Screen
22
Complexity of Rendering Scattering Media
Objects
Virtual Viewpoint
Virtual Screen
23
Complexity of Rendering Scattering Media
Objects
Virtual Viewpoint
Virtual Screen
24
Complexity of Rendering Scattering Media
Objects
Virtual Viewpoint
Virtual Screen
640 x 480 (image) x 4 (lights) x 50 (steps)
100 ( directions ) x 50 (steps) x 30
(intersect) ?
1.9 Trillion Calculations 3.0 GHz
CPU?
25
The amount of contributing light is equal to the
area of fog between the camera and the point in
the screen. The light of the incoming ray,
however, was partially absorbed by the fog
itself, thus reducing its intensity.
Amount of Light absorbed by fog
Area of fog absorbing light
Area of fog emitting light
Intensity of the light coming from the fog
Intensity of the light coming from the scene
Amount of Light emitted by fog
26
Since the area of fog emitting light is the same
for the area of fog absorbing light, and the
assumption is made that the amount of light
emitted is the same percentage as absorbed, then
this equal simplifies to
Area of fog
Intensity of the light coming from the scene
Intensity of the light coming from the fog
Amount of light absorbed/emitted by fog (fog
density)
27
Radiation Fog
Advection Fog
Simulated Foggy Image
Actual Clear Day Image
Simulated Foggy Image
Actual Clear Day Image
Dense Aerosols with Drizzle
Haze
Simulated Hazy Image
Actual Clear Day Image
Simulated Foggy Image
Actual Clear Day Image
Urban Aerosol with Moderate Rain
Fog with Cumulus Clouds
Simulated Foggy Image
Simulated Foggy Image
Actual Clear Day Image
Actual Clear Day Image
28
The Fundamental Assumption in Vision
Lighting
No Change in Radiance
Surface
Camera
Assumption We live in Vacuum!
29
Driving in Bad Weather
People tend to drive fast in fog!! Nature, 1998
30
Attenuation Model Zeroth Order Scattering
Scattering Medium
Attenuated Exiting Light
Incident Light
X d
Unit Cross Section
dx
X 0
Brightness at Distance d
( Bouguers Law, 1729 )
31
Airlight Model First Order Scattering
( Koschmeider, 1924 )
Sunlight
Diffuse Skylight
dV
Observer
Object
d
Diffuse Ground Light
Brightness due to a Path of Length d
Horizon Brightness
32
Distant objects appear Bright !
Mountains
33
Structure from Airlight
34
How does Brightness/Color vary with Distance?
Object
Observer
d
Color
Color
Distance
Distance
Attenuation
Airlight
35
Contrast Degradation in Bad Weather
Irradiance Attenuation
Airlight


Scattering Coefficient
Reflectance
Horizon Brightness
Depth
(1)
(2)
Contrast between Iso-Depth points , P
and P
Contrast Decay Exponential in Scene Depth
36
Depth Edges vs. Reflectance Edges
Mild Fog
Denser Fog
Reflectance Edge
Depth Edge
Normalized SSD of Depth Edge Neighborhood
Normalized SSD of Reflectance Edge
Neighborhood
37
Edge Classification from Weather Changes
Edge Classification
Mild Fog
Denser Fog
Reflectance Edge Depth Edge
38
Defogging Videos
Foggy Video
Defogged Video
Histogram Equalized Video
39
Scene Structure from Weather Changes
Irradiance under versus
Irradiance under Linear
All Scene points at Depth 1
All Scene points at Depth 2
Scaled Depth
40
Gray World Contrast Restoration and Structure
3D Visualization
Deweathering
41
Contrast Restoration and 3D Structure
Dense Fog, 530 PM
Mild Fog, 5 PM
Contrast Restored Image
Computed Depth Map (20 levels)
42
Scattering and Wavelength
constant
Rayleighs Law
(0 4)
Smaller the particles, larger the dependence on
wavelength Blue skies through pure air (small
particles) Fog looks greyish (whitish) larger
water droplets.
43
Clear Day from Hazy Days
Unknown Hazy Conditions
Time 300 PM
Time 330 PM
( Narasimhan et. al, IJCV 2002)
44
Clear Day from Hazy Day Using Polarizing Filters
De-hazed
Haze
Airlight is Partially Polarized
45
Scattering from Near-Field Sources
46
Scattering from Near-Field Sources
Loss of contrast
47
Scattering from Near-Field Sources
Dimming and blur
Loss of contrast
48
Scattering from Near-Field Sources
Glows
Dimming and blur
Lost of contrast
49
(No Transcript)
50
Other 3D shape cues

• Shading

Merle Norman Cosmetics, Los Angeles
51
Surface normals in C orthographic projection
52
Constraint on surface normals
53
How to use the constraints?
54
(No Transcript)
55
(No Transcript)
56

• Shading
• Texture

The Visual Cliff, by William Vandivert, 1960
57

• Shading
• Texture
• Focus

From The Art of Photography, Canon
58
Active Methods
Li Zhangs one-shot stereo

• Project structured light patterns onto the
  object
• simplifies the correspondence problem

59
Active Stereo with Structured Light
60
Laser Scanning
Digital Michelangelo Project http//graphics.stanf
ord.edu/projects/mich/

• Optical triangulation
• Project a single stripe of laser light
• Scan it across the surface of the object
• This is a very precise version of structured
  light scanning

61
Laser Scanned Models
The Digital Michelangelo Project, Levoy et al.
62
Laser Scanned Models
The Digital Michelangelo Project, Levoy et al.
63
Laser Scanned Models
The Digital Michelangelo Project, Levoy et al.
64
Laser Scanned Models
The Digital Michelangelo Project, Levoy et al.