Page 61 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

38 Chapter 2. Video Coding: Fundamentals

2.6.6 Other Coding Methods
There are many other intraframe coding techniques. Examples are block-
truncation coding, fractal coding, quad-tree and recursive coding, multireso-
lution coding, and neural-network-based coding. A detailed (or even a brief)
discussion of such techniques is beyond the scope of this book, and the
interested reader is referred to Ref. 52.

2.7 Interframe Coding

As already discussed, video is a time sequence of still images or frames. Thus,
a naive approach to video coding would be to employ any of the still-image
(or intraframe) coding methods discussed in Section 2.6 on a frame-by-frame
basis. However, the compression that can be achieved by this approach is
limited because it does not exploit the high temporal correlation between the
frames of a video sequence. Interframe coding refers to video coding tech-
niques that achieve compression by reducing this temporal redundancy. For
this reason, such methods are also known as temporal redundancyreduc-
tion techniques. Note that interframe coding may not be appropriate for some
applications. For example, it would be necessary to decode the complete inter-
frame coded sequence before being able to randomly access individual frames.
Thus, a combined approach is normally used in which a number of frames
are intraframe coded (I-frames) at speci c intervals within the sequence and
the other frames are interframe coded (predicted or P-frames) with reference
to those anchor frames. In fact, some systems switch between interframe and
intraframe within the same frame.

2.7.1 Three-Dimensional Coding
The simplest way to extend intraframe image coding methods to interframe
video coding is to consider 3-D waveform coding. For example, in 3-D trans-
form coding based on the DCT, the video is rst divided into blocks of
M × N × K pels (M; N; K denote the horizontal, vertical, and temporal dimen-
sions, respectively). A 3-D DCT is then applied to each block, followed by
quantization and symbol encoding, as illustrated in Figure 2.14. A 3-D coding
method has the advantage that it does not require the computationally intensive
process of motion estimation (as will be discussed in Section 2.7.2). However,
it requires K frame memories both at the encoder and decoder to bu6er the
frames. In addition to this storage requirement, the bu6ering process limits the
use of this method in real-time applications because encoding=decoding cannot
begin until all of the next K frames are available. In practical systems, K is
typically set to 2– 4 frames.