Motion in Computer Vision

blackeningfourAI and Robotics

Oct 19, 2013 (3 years and 9 months ago)

102 views

Motion in Computer Vision

E. Scott Larsen

October 29, 2001

Comp 256

Motion in Computer Vision


The Primary Assumption


The BCCE


Space
-
time Images


Examples


The Aperture Problem


The Correspondence Problem (again!)


Optical Flow

The Primary Assumption


Observed brightness of any object is constant over
time [2]


Motion and gray value changes correspond to
each other [1] (in gray images)

The BCCE

BCCE: Brightness Change Constraint Equation

Space
-
time Images

Examples

Examples

Examples

Examples

Examples

Examples

Conclusion From the Examples


The Primary Assumption is not always valid (we
already knew that)




But… there are other issues also…

The Aperture Problem

From left to right, the line appears to have moved in
the direction indicated. But…

The Aperture Problem

…We open our eyes a little farther, and find that we
were wrong.

The Correspondence Problem

The grand
-
pappy of vision problems attacks again!

Optical Flow


“The
optical flow

is a vector field subject to the
BCCE, and is loosely defined as the
apparent
motion

of the image brightness pattern.”[3]


It originates from fluid dynamics. [1]


It is not the same as the motion field: “projection
of 3
-
D physical motion field by the optics onto the
image plane.” [1]

Summary


The primary assumption is that image changes and
motion directly correspond to each other


Point

point correspondences are required


Motion is currently much more useful than I suggest,
and there is much room for original investigation.


References


[1] J
ä
hne B.,
Digital Image Processing
, 1997,
Springer
-
Verlag


[2] Sonka M., V. Hlavac and R. Boyle,
Image
Processing, Analysis, and Machine Vision
, 1999,
Brooks/Cole


[3] Trucco E., and A. Verri,
Introductory
Techniques for 3
-
D Computer Vision
, 1998,
Prentice
-
Hall