Page 117 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

94 Chapter 4. Basic Motion Estimation Techniques

frame. The prediction error, called the displaced-frame di erence (DFD), is
encoded instead of the current frame itself. The estimated motion information
also has to be transmitted, unless the decoder can estimate it from previously
decoded information. This section introduces the basics of motion estimation.
It de nes and formulates the motion estimation problem and describes the
main approaches and models used to solve this problem. Examples of such
solutions will be discussed in subsequent sections.

4.2.1 Projected Motion and Apparent Motion

In video, the 3-D motion of objects in space is projected as 2-D motion onto
the image plane. This 2-D motion, called projected motion, is illustrated in
Figure 4.1. Thus, motion estimation may refer to the process of estimating
image-plane 2-D motion or object-space 3-D motion. Note that the two are
not equivalent. In fact, 2-D motion estimation is usually the rst step toward
3-D motion estimation. This chapter considers 2-D motion estimation only.
For 3-D motion estimation, the reader is referred to Ref. 10.
In video coding, motion is estimated by observing the spatiotemporal vari-
ation of intensity between frames. This is called the apparent motion.Inthe
ideal case, apparent motion is equivalent to true projected motion. In practice,
however, this is not always the case. For example, when a circle with uniform

Y

object space
y
2-D projected
X
motion O
p'
P'
center of
p prejection
3-D
motion
P x
image plane
Z

Figure 4.1: Projected motion