Page 28 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Section 1.3. Possible Solutions 5

the displaced-frame di*erence (DFD). Instead of encoding the current frame
itself, this error signal is encoded, since it has a much reduced entropy. At
the decoder, the same reference frame is used along with the received motion
information to produce the same prediction. This prediction is then added to
the received error signal to reconstruct the current frame.
Careful examination of this codec (as will be detailed in subsequent chap-
ters) reveals that a motion-based approach can be adopted to provide suitable
solutions for the three challenges of higher coding e(ciency, reduced com-
plexity, and error resilience. This motion-based approach can be summarized
as follows:

1. Advanced motion estimation techniques. One way to achieve higher cod-
ing e(ciency is to improve the performance of the motion estimation
and compensation processes. The aim is to produce a better motion-
compensated prediction and consequently reduce the entropy of the DFD
signal. This should be achieved at the same or, preferably, a reduced
motion overhead. 4
2. Reduced-complexity motion estimation techniques. Motion estimation is
the most computationally intensive process in a typical video codec. In
fact, pro1ling results (as will be shown in Chapter 7) indicate that the
computational complexity of this process is greater than that of all the
remaining encoding steps combined. Thus, by reducing the complexity
of this process, the overall complexity of the codec can be reduced.
3. Motion-compensated error concealment techniques. Apart from control
and header data, the output of a typical video codec is one of two types:
5
motion data or error (i.e., DFD) data. Among the two types, motion data
carries, in general, most of the information about a frame. In fact, at
very low bit rates (typical of mobile video communication), motion data
consumes a very high percentage of the available bit budget [8]. Thus,
in the case of errors, it is very important to recover lost or erroneously
received motion information. A class of error-resilience techniques that
achieves this is motion-compensated error concealment, also known as
temporal error concealment. Such techniques are particularly suited for
mobile video communication, since, unlike other error resilience tech-
niques, they do not increase the bit rate and they do not introduce any
delay.

4 An increase in motion overhead can be tolerated provided that the overall rate-distortion
performance is improved.
5 In the case of intracoded frames, the error signal is the same as the frame signal and no
motion data is transmitted.