P1: IML/FFX P2: IML/FFX QC: IML/FFX T1: IML
Video ̇Compression ̇OLE WL040/Bidgolio-Vol I WL040-Sample.cls September 14, 2003 18:10 Char Count= 0
Video CompressionVideo Compression
Immanuel Freedman,Dr. Immanuel Freedman, Inc.Introduction 537
Definition of Video Compression 537
Example of Video Compression 537
Objective of Video Compression 537
Compression Paradigms 537
Operations and Infrastructure Supported
by Digital Video 539
Digital Video Signal Representation 539
Anatomy of a Digital Video Sequence 539
Sampling, Quantizing, and Coding 540
Time Code 541
Digital Video Signal Compression 542
Overview and Tradeoff Dimensions 542
Digital Video Quality Assessment 542
Rate Distortion Relationships 542
Principles of Digital Video Compression 543
Pre- and Postprocessing 543
Motion Estimation and Compensation 543
Rate Control of Compressed Digital Video 544Psychovisual Modeling 544
Statistical Multiplexing 544
Network Transmission Issues 544
Digital Video Compression Standards 544
MPEG-1, -2, -4 Visual Codecs 544
MPEG-7 Visual and MPEG-21 Standards 545
ITU-T Visual Codecs 545
Proprietary Codecs 547
How Standards Are Defined and Described 548
Digital Video Application Solutions 550
Digital Video Business Models 550
Video-on-Demand over Broadband Networks 550
Customer Relationship Management over
Third Generation Mobile Networks 551
Conclusion and Future Outlook 551
Glossary 551
Cross References 551
References 551INTRODUCTION
Definition of Video Compression
Video compression is a process of reducing the amount
of digital data used to represent a sequence of images
normally varying in time and intended to portray motion
subject to the requirements that the quality of the recon-
structed video is sufficient for a certain application and
the complexity of the computation involved is appropri-
ate for that application.
Some researchers, for example, Shi and Sun (2000),
use the termvideoto refer exclusively to image frames
and sequences associated with the visible band of the elec-
tromagnetic spectrum while others, for example, Tekalp
(1996), refer exclusively to sequences.Example of Video Compression
Figure 1a shows images of frames numbered 1231–1233
from “The Emotion of Space” movie (NASA, 2001) indi-
cating a scene change at frame 1232 and motion sequence
from frames 1232 to 1233. Figure 1b displays a graph of
the data rate and sample size by frame number centered
on frame 1231 for the aforementioned movie. With an im-
age size of 320×240 pixels, pixel depth of 24 bits (8 for
each of the red, green, and blue channels), and an aver-
age frame rate of 15 frames per second (fps) to create an
illusion of motion by the persistence of vision, the 3219
frames of this movie of about 215 s duration would re-
quire a data rate of about 27.65 Mbps to be transmitted in
real time without compression, not considering the audio
tracks, far above even the capacity of a 1.5 Mbps T1 com-
munications line with a download time for the 742 MB
data exceeding 8 min under optimal conditions.Objective of Video Compression
Video compression provides a technology solution for ap-
plications in which the required data rate for communica-
tion, manipulation, or storage of digital video sequences
exceeds the capacity of communications channels or stor-
age devices. Table 1 lists the data rate and correspond-
ing compression ratio required for several applications in
common use.Compression Paradigms
Video compression methods are often described as
“lossless” or “lossy.” Lossless compression methods
are reversible unlike lossy methods, which comprise
irreversible changes to data. The term lossy arose from
an analogy with a concept familiar to electrical engineers,
the dissipation of electrical energy as heat in the trans-
mission of power. The approximations to data made by
lossy compression methods frequently yield far greater
compression than the exact lossless methods for a small
reduction in information content. In hybrid schemes,
residuals from the approximations made by a high com-
pression (perhaps 100:1) lossy method can, in turn, be
coded by a moderate compression (perhaps 2:1) loss-
less method to yield an exactly reversible combination of
compression methods. This allows the user to browse a
highly compressed approximation to the data and make
precise calculations with an exact reconstruction when
required. The terms “approximate” and “exact” may be
more acceptable than the terms lossy and lossless to sci-
entists and administrators, who often take the stance
that “loss is intolerable” but “an approximation is good
enough.”537