Optical Flow, KLT Feature Tracker - Yonsei University · 2014-12-29 · E-mail:...
Transcript of Optical Flow, KLT Feature Tracker - Yonsei University · 2014-12-29 · E-mail:...
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Optical Flow,Optical Flow,KLT Feature TrackerKLT Feature Tracker
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion in Computer VisionMotion in Computer Vision
Motion
• Structure from motion• Detection/segmentation with direction
[1]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field Motion Field v.sv.s. Optical Flow [2], [3]. Optical Flow [2], [3]
Motion Field:
an ideal representation of 3D motion as it is projected onto
a camera image.
Optical Flow: the approximation (or estimate) of the motion field
which can
be computed from time-varying image sequences. Under the simplifying
assumptions of 1) Lambertian
surface, 2) pointwise
light source at infinity,
and 3) no photometric distortion.
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field [2]Motion Field [2]
•
An ideal representation of 3D motion as it is projected onto a camera
image.
•
The time derivative of the image position of all image points given that
they correspond to fixed 3D points. “field : position
vector”
•
The motion field v
is defined as
whereP
is a point in the scene where Z is the distance to that scene point.
V
is the relative motion between the camera and the scene,
T
is the translational component of the motion,
and ω is the angular velocity of the motion.
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field [2]Motion Field [2]
Zf Pp (1)
3D point P
(X,Y,Z) and 2D point p
(x,y), focal length f
Motion field v
can be obtained by taking the time derivative of (1)
ZYfyZXfx
2
2
ZVY
ZVfv
ZVX
ZVfv
ZYy
ZXx
(2)
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field [2]Motion Field [2]
PTV
(3)
The motion of 3D point P, V
is defined as flow
By substituting (3) into (2), the basic equations of the motion field is acquired
fy
fxyxf
ZfTyTv
fx
fxyyf
ZfTxTv
XYZX
YZy
YXZY
XZx
2
2
(4)
X X Y Z
Y Y Z X
Z Z X Y
V T Z YV T X ZV T Y X
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field [2]Motion Field [2]
The motion field is the sum of two components, one of which depends on translation only, the other on rotation only.
fy
fxyxf
ZfTyTv
fx
fxyyf
ZfTxTv
XYZX
YZy
YXZY
XZx
2
2
ZfTyTv
ZfTxTv
YZy
XZx
fy
fxyxfv
fx
fxyyfv
XYZXy
YXZYx
2
2
Translational components
Rotational components
(4)
(5) (6)
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field: Pure Translation [2]Motion Field: Pure Translation [2]
If there is no rotational motion, the resulting motion field has
a peculiar spatial
structure.
If (5) is regarded as a function of 2D point position,
ZfTyTv
ZfTxTv
YZy
XZx
(5)
Z
YZy
Z
XZx
TTfy
ZTv
TTfx
ZTv
Z
Y
Z
X
TTfy
TTfx
0
0
If (x0
, y0
) is defined as in (6)
0
0
yyZTv
xxZTv
Zy
Zx(6) (7)
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field: Pure Translation [2]Motion Field: Pure Translation [2]
Equation (7) say that the motion field of a pure translation is radial. In particular, if TZ
<0, the vectors point away from p0 (x0
, y0
), which is called the focus of expansion (FOE). If TZ
>0, the motion field vectors point towards p0
, which is called the focus of contraction. If TZ
=0, from (5), all the motion field vectors are parallel.
ZfTyTv
ZfTxTv
YZy
XZx (5)
ZTfvZ
Tfv
Yy
Xx
(8)
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field: Motion Parallax [6]Motion Field: Motion Parallax [6]
Equation (8) say that their lengths are inversely proportional to the depth of the corresponding 3D points.
ZTfvZ
Tfv
Yy
Xx
(8)
This animation is an example of parallax. As the viewpoint moves
side to side, the objects in the distance appear to move more slowly than the objects close to the camera [6].
http://upload.wikimedia.org
/wikipedia/commons/a/ab/
Parallax.gif
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field: Motion Parallax [2]Motion Field: Motion Parallax [2]
If two 3D points are projected into one image point, that is coincident, rotational component will be the same. Notice that the motion vector V
is about camera
motion.
(4)
fy
fxyxf
ZfTyTv
fx
fxyyf
ZfTxTv
XYZX
YZy
YXZY
XZx
2
2
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field: Motion Parallax [2]Motion Field: Motion Parallax [2]
The difference of two points’ motion field will be related with translation
components. And, they will be radial w.r.t
FOE or FOC.
01 2 1 2
01 2 1 2
1 1 1 1
1 1 1 1
x Z X Z
y Z Y Z
v T x T f x x TZ Z Z Z
v T y T f y y TZ Z Z Z
FOC
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field: Motion Parallax [2]Motion Field: Motion Parallax [2]
Motion ParallaxMotion Parallax
The relative motion field of two instantaneously coincident points:
1.
Does not depend on the rotational component of motion
2.
Points towards (away from) the point p0, the vanishing point of the translation direction.
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field: Pure Rotation Motion Field: Pure Rotation w.r.tw.r.t
YY--axis [7]axis [7]
If there is no translation motion and rotation w.r.t
x-
and z-
axis, from (4)
fxyv
fxfv
Yy
YYx
2
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Motion Field: Pure Rotation Motion Field: Pure Rotation w.r.tw.r.t
YY--axis [7]axis [7]
Z
Distance to the point, Z, is constant.
Translational Motion
Rotational Motion
Z
Distance to the point, Z, is changing. According to Z, y is changing, too.
Z1
Z2
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Estimation of the Optical Flow [4]Estimation of the Optical Flow [4]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
The Image Brightness Constancy Equation [2]
Estimation of the Optical Flow [4]Estimation of the Optical Flow [4]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
tT II V
I
tIV
AssumptionThe image brightness is continuous and differentiable as many times as needed in both the spatial and temporal domain.
The image brightness can be regarded as a plane in a small area.
Estimation of the Optical Flow [4]Estimation of the Optical Flow [4]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Optical Field: Aperture Problem [2], [4], [9]Optical Field: Aperture Problem [2], [4], [9]
The component of the motion field in the direction orthogonal to
the spatial image
gradient is not constrained by the image brightness constancy equation.
Given
local information can determine component of optical flow vector
only in direction
of brightness gradient.
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Optical Field: Aperture Problem [2], [9]Optical Field: Aperture Problem [2], [9]
The aperture problem. The grating
appears to be moving down and to the right, perpendicular
to the orientation of the bars. But it could be moving in many other directions, such as only down, or only to the right. It is impossible to determine unless the ends of the bars become visible in the aperture.
http://upload.wikimedia.org/wikipedia/commons/f/f0/Aperture_pr
oblem_animated.gif
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Optical Field: Methods for Determining Optical Flow [4]Optical Field: Methods for Determining Optical Flow [4]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Optical Field: Phase Correlation Method [10]Optical Field: Phase Correlation Method [10]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Optical Field: Phase Correlation Method [10]Optical Field: Phase Correlation Method [10]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Optical Field: Phase Correlation Method [10]Optical Field: Phase Correlation Method [10]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Optical Field: LucasOptical Field: Lucas--KanadeKanade
Method [8]Method [8]
Assuming that the optical flow (Vx
,Yy
) is constant in a small window of size mxm
with m>1, which is center at (x, y) and numbering the pixels within as 1…n, n=m2, a set of equations can be found:
tT II V
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
Optical Field: LucasOptical Field: Lucas--KanadeKanade
Method [8]Method [8]
cf.) Harris corner detector
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
KLT Feature Tracker [11]KLT Feature Tracker [11]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
KLT Feature Tracker [11]KLT Feature Tracker [11]
F’(x)로 weighting
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
KLT Feature Tracker [11]KLT Feature Tracker [11]
KLT: An Implementation of the Kanade-Lucas-Tomasi
Feature Tracker
http://www.ces.clemson.edu/~stb/klt/
tT II V
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
LucasLucas--KanadeKanade
Method: A Unifying Framework [12]Method: A Unifying Framework [12]
Since the Lucas-Kanade algorithm was proposed in 1981 image alignment has becomeone of the most widely used techniques in computer vision.
Besides optical flow, some of its other applications include- tracking (Black and Jepson, 1998; Hager and Belhumeur, 1998),- parametric and layered motion estimation (Bergen et al., 1992),- mosaic construction (Shum and Szeliski, 2000),- medical image registration (Christensen and Johnson, 2001), - face coding (Baker and Matthews, 2001; Cootes et al., 1998).
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
LucasLucas--KanadeKanade
Method: A Unifying Framework [12]Method: A Unifying Framework [12]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
LucasLucas--KanadeKanade
Method: A Unifying Framework [12]Method: A Unifying Framework [12]
Minimizing the expression in Equation (1) is a non-linear optimization task even if W(x; p) is linear in p because the pixel values I(x) are, in general, non-linear in x.
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
LucasLucas--KanadeKanade
Method: A Unifying Framework [12]Method: A Unifying Framework [12]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
LucasLucas--KanadeKanade
Method: A Unifying Framework [12]Method: A Unifying Framework [12]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
LucasLucas--KanadeKanade
Method: A Unifying Framework [12]Method: A Unifying Framework [12]
Setting this expression to equal zero and solving gives the closed form solution for the minimum of the expression in Equation (6) as:
SD
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
LucasLucas--KanadeKanade
Method: A Unifying Framework [12]Method: A Unifying Framework [12]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
LucasLucas--KanadeKanade
Method: A Unifying Framework [12]Method: A Unifying Framework [12]
E-mail: [email protected]://web.yonsei.ac.kr/hgjung
ReferencesReferences
1. Richard Szeliski, “ Dense motion estimation,” Computer Vision: Algorithms and Applications, 19 June 2009 (draft), pp. 383-426.
2. Emanuele Trucco, Alessandro Verri, “8. Motion,” Introductory Techniques for 3-D Computer Vision, Prentice Hall, New Jersey 1998, pp.177-218.
3. Wikipedia, “Motion field,” available on www.wikipedia.org.4. Wikipedia, “Optical flow,” available on www.wikipedia.org.5. Alessandro Verri, Emanuele Trucco, “Finding the Epipole from Uncalibrated Optical Flow,”
BMVC 1997, available on http://www.bmva.ac.uk/bmvc/1997/papers/052/bmvc.html.6. Wikipedia, “Parallex,” available on www.wikipedia.org.7. Jae Kyu Suhr, Ho Gi Jung, Kwanghyuk Bae, Jaihie Kim, “Outlier rejection for cameras on
intelligent vehicles,” Pattern Recognition Letters 29 (2008) 828-840.8. Wikipedia, “Lucas-Kanade Optical Flow Method,” available on www.wikipedia.org.9. Wikipedia, “Aperture Problem,” available on www.wikipedia.org.10. Wikipedia, “Phase correlation,” available on http://en.wikipedia.org/wiki/Phase_correlation11. Wikipedia, “Kanade-Lucas-Tomasi feature tracker,” available on
http://en.wikipedia.org/wiki/Kanade%E2%80%93Lucas%E2%80%93Tomasi_feature_track er
12. Simon Baker, Iain Matthews, “Lucas-Kanade 20 Years: A Unifying Framework,” International Journal of Computer Vision, 56(3), 221-255, 2004.