Page 26 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Section 1.2. Main Challenges 3

by video, the use of e(cient coding techniques is vital. For example,
2
real-time transmission of a CIF video at 15 frames=s over a 9:6 kbits=s
GSM channel requires a compression ratio of about 1900:1. Although
current coding techniques are capable of providing such compression
ratios, there is a need for even higher coding e(ciency to improve
the quality (i.e., larger formats, higher frame rates, and better visual
quality) of video at such very low bit rates. This continues to be the
case even with the introduction of enhancements to second-generation
systems, like the General Packet Radio Service (GPRS) [3] and the
Enhanced Data Rates for GSM Evolution (EDGE), and also with the
deployment of future higher-capacity, third-generation systems, like the
Universal Mobile Telecommunication System (UMTS) [4].

2. Reduced computational complexity. In mobile terminals, processing
power and battery life are very limited and scarce resources. Given the
signi1cant amount of computational power required to process video, the
use of reduced-complexity techniques is essential. For example, recent
implementations of video codecs [5,6] indicate that even state-of-the-art
digital signal processors (DSPs) cannot, yet, achieve real-time video en-
coding. Typical results quoted in Refs. 5 and 6 are 1–5 frames=s using
small video formats like SQCIF and QCIF. 3
3. Improved error resilience. The mobile channel is a hostile environment
with high bit error rates caused by a number of loss mechanisms, like
multipath fading, shadowing, and co-channel interference. In the case of
video, the eBects of such errors are magni1ed due to the fact that the
video bitstream is highly compressed to meet the stringent bandwidth
limitations. In fact, the higher the compression is, the more sensitive
the bitstream is to errors, since in this case each bit represents a larger
amount of decoded video. The eBects of errors on video are also magni-
1ed by the use of predictive coding and variable-length coding (VLC).
The use of such coding methods can lead to temporal and spatial error
propagation. It is, therefore, not di(cult to realize that when transmitted
over a mobile channel, compressed video can suBer severe degradation
and the use of error-resilience techniques is vital.

2 CIF stands for Common Intermediate Format. It is a digital video format in which the
luminance component is represented by 352 pels × 288 lines and the two chrominance components
are each of dimensions 176 × 144, where each pel is usually represented by 8 bits. Digital video
formats are discussed in more detail in Chapter 2.
3 Quarter-CIF (QCIF) has a luminance component of 176 × 144, whereas sub-QCIF (SQCIF)
has a luminance component of 128 × 96.