스테레오 정합, 신뢰 전파 기법에 대한 개념과 간단한 예제.
[참고]
J.H. Kim, and Y.H. Ko, “Multibaseline based Stereo Matching Using Texture adaptive Belief Propagation Technique." Journal of the Institute of Electronics and Information Engineers Vol. 50, No. 1, pp.75-85, 2013.
3. 2018-10-05
Introduction
3
Stereo Vision
: Stereo vision is the extraction of 3D information from digital images, such as obtained by a CCD camera. By comparing
information about a scene from two vantage points, 3D information can be extracted by examination of the relative
positions of objects in the two panels. This is similar to the biological process Stereopsis.
4. 2018-10-05
Introduction
4
Applications
: Stereo vision is highly important in fields such as robotics, to extract information about the relative position of 3D objects in the
vicinity of autonomous systems.
6. 2018-10-05
Camera Geometry
6
Pin-hole camera model
Z
X
C
x'
p
x
xf
x
XZ plane
x
XxZfx ::
ZxXfx
Z
Xf
x x
Z
Y
C
x'
p
y
yf
x
YZ plane
y
YyZf y ::
ZyYf y
Z
Yf
y
y
Z
Y
X
Z)Y,(X,x
C
x'
p
x
y
x
x
p
Z
Xf
y
y
p
Z
Yf
Image sensor
Pin-hole
),( yx ppPrincipal point
cf. Principal point
)(x,y,z
8. 2018-10-05
Stereo Matching
8
lPMC ll LCP
f
x
Z
X l
f
x
Z
bX r
Z
f
x
X l
bZ
f
xx
bZ
f
x
Z
f
x rlrl
,
rl xx
bf
Z
bZ
f
x
X r
),,( ZYXP
Z
f
lx rx
),( lll yxP ),( rrr yxP
b
rClC
Left camera
lens center
Right camera
lens center
RightCameraOpticalAxis
LeftCameraOpticalAxis
Focal Length
Object Point
M N
L R
Stereo Camera model
similar
similar
rPNC rr RCP
from a from b
a
b
Left image plane Right Image Plane
We need to disparity
information
11. 2018-10-05
Stereo Matching
11
Local method
SSD (Sum of Squared Difference)
M
y
N
x
rlMN ydxIyxIdyxSAD
1 1
|),(),,(|),,(
M
y
N
x
rlMN ydxIyxIdyxSSD
1 1
2
),(),,(),,(
M
y
N
x
rlMN ydxIyxI
NM
dyxMAE
1 1
|),(),,(|
1
),,(
SAD (Sum of Absolute Difference)
MAE(Mean Absolute Error)
12. 2018-10-05
Stereo Matching
12
Global method
: Use the energy functions
)()()( dEdEdE smoothnessdata
Belief Propagation(BP), Graph Cut(GC), Dynamic Programing(DP)…
Implementation :
16. 2018-10-05
Belief Propagation
16
Likelihood
s
sdsFXYP ,exp| sd
Matching cost function
YP
XPXYP
YXP
|
)|(
: disparity candidate at pixel s
|),(),(|, jdiIjiIdsF srls
Example
Disparity(𝑥 𝑠)20 25 117 135 150 118 134 22 20 15 d=0
Left Image(Ref.) Right Image
d=1 d=2 d=3 d=4
Value(𝑦𝑠) 135 130 128 16 32
17. 2018-10-05
Belief Propagation
17
Prior
sd
Constraint function
YP
XPXYP
YXP
|
)|(
: disparity candidate at pixel s
t
t
st t
s sNt
ts ddVXP
)(
,exp
tsts ddddV ,
Example
t
t
st t
: disparity candidate at pixel ttd
18. 2018-10-05
Belief Propagation
18
Final model
XPXYP |
s sNt
ts
s
s ddVdsF ,exp,exp
YP
XPXYP
YXP
|
)|(
s sNt
tsst
s
sss xxyx ,,
st
MAP maximize marginal probability (maximize belief)
s : local evidence for node
: compatibility between nodes
sx
sx and ty
t
t
st t
21. 2018-10-05
Belief Propagation
21
Implementation of message, belief and disparity
tsk
s
xxNx
s
i
kstsstsss
x
t
i
st xmxxyxxm
)(
1
)(),(),(max)(
tsk xxNx
skssssss xmyxxb
)(
)(),()(
)(maxarg ks
x
MAP
s xbd
k
jik
i
xxNx
i
t
kijicii
x
j
t
ij xMxxxMcxM
)(
1
)(),()(min)(
jik xxNx
ikiiiii xMxMcxB
)(
)()()(
)(minarg ks
x
MAP
s xBx
k
take the negative logarithm of each equation