7. Introduction(3)
Proposed method(3/7)
Experiment(6)
Discussion(2)
CG Lab, Hiroshima City University
http://www.cg.info.hiroshima-cu.ac.jp/
Sun
Unpolarized
Scattering
Partially
polarized
light
Component parallel to scattering plane (superscript: || )
Component perpendicular to scattering plane (superscript:
Polarization of scattered light
)
8. Introduction(3)
I
I
||
Proposed method(4/7)
||
A 1
A 1
e
e
Z
1
Z
2
1
Experiment(6)
Re
Z
CG Lab, Hiroshima City University
http://www.cg.info.hiroshima-cu.ac.jp/
Z
Re
Discussion(2)
Light
source
2
Haze
Maximum scattered light A
Distance
Camera
Observed Scattered light A
light
Attenuated object light T
I
Attenuation exp(- Z)
Distance Z
Object
Object light R
Camera
Formulation of observed light
10. Introduction(3)
Proposed method(6/7)
Observed light
I 1p
Camera
Distance Z1
Experiment(6)
CG Lab, Hiroshima City University
http://www.cg.info.hiroshima-cu.ac.jp/
Discussion(2)
Haze
Input:
Traffic
sign 1
Traffic
sign 2
Observed light
I 2q
||
A ,A ,
1
P
||
||
I1 p
A 1
e
Z1
A 1
e
Z1
1
2
||
2q
A 1
I 2q
A 1
I
1
2
p P
I1 p
Traffic
sign 2
Object light
R 2q
arg min
||
Q
Traffic
sign 1
||
A , A ,
A ,A ,
1
Camera
||
Output:
Distance Z 2
Object light
R 1p
||
I 1 , I 1 , I 2 , I 2 , Z 1 , Z 2 , R1 , R 2
||
e
Z2
1
2
q Q
e
Z2
1
2
2
R1 p e
Z1
2
R1 p e
Z1
2
R2qe
Z2
R2q e
Z2
2
Levenberg-Marquardt method
A
||
64 for 8bit camera
A
[Initial value]
1
Z1
Z2
2
Parameter estimation from two references
11. Introduction(3)
Proposed method(7/7)
Experiment(6)
Discussion(2)
CG Lab, Hiroshima City University
http://www.cg.info.hiroshima-cu.ac.jp/
[Schechner 2003] use sky region as A
A
Reason 1: Stratosphere is far
Reason 2: Universe is dark
If sky is unobserved...
If mountain boundary undetected...
Related work
18. Introduction(3)
Proposed method(7)
Experiment(6)
Discussion(1/2)
CG Lab, Hiroshima City University
http://www.cg.info.hiroshima-cu.ac.jp/
Sky area not concerned
Image enhanced at
not only reference objects
but also other objects
Close objects fail
Particle distribution isn’t uniform
Particle size isn’t same
Water tank size is finite
Specular reflection of reference object
Dark diffuse reflection
Illumination isn’t uniform
Close distance from illumination
Polarization of water surface
Affected by incident angle
Discussion
19. Introduction(3)
Proposed method(7)
Experiment(6)
Discussion(2/2)
CG Lab, Hiroshima City University
http://www.cg.info.hiroshima-cu.ac.jp/
Color relatime polarization camera
Set camera on vehicles
Traffic sign recognition
On-line parameter updation
High precision using 3 or more traffic signs
Creating traffic sign database
Compute distance from traffic sign size
Intrinsic camera calibration
Future work
20. (c) Daisuke Miyazaki 2013
All rights reserved.
http://www.cg.info.hiroshima-cu.ac.jp/~miyazaki/
Daisuke Miyazaki, Daisuke Akiyama, Masashi Baba, Ryo
Furukawa, Shinsaku Hiura, Naoki Asada, “Polarization-based
dehazing using two reference objects,” CPCV, 2013.
Computer Graphics Laboratory, Hiroshima City University
http://www.cg.info.hiroshima-cu.ac.jp/