The Internet Encyclopedia (Volume 3)

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

544 VIDEOCOMPRESSION

together with the estimate of their prediction error (Shi

& Sun, 2000). The three-dimensional transform tech-

nique has become feasible due to recent advances in elec-

tronics manufacture and represents a growing area for

research.

Rate Control of Compressed Digital Video

The information content of digital video sources may vary

in complexity from scene to scene. Constant bit rate en-

coding forces the reconstructed video quality to vary in or-

der to maintain a (near) constant bit rate for the channel

coder. Variable bit rate (VBR) has the goal of maintaining

(near) constant quality by varying the bit rate allocated

to scenes of varying complexity. Popular methods of rate

control include varying the quantization of transform cod-

ing in a feedback loop and embedded zero tree encoding

(Shapiro, 1993).

Psychovisual Modeling

Human attention is often attracted to faces and facial

expressions and distracted by the motion of peripheral

objects across the field of view (see Colmenarez, Frey, &

Huang, 1999). An encoder may choose to encode these

regions of interest at a higher bit rate than others if the

regions can be bounded (typically, by focus boxes).

Statistical Multiplexing

Often, a broadcaster wishes to transmit several program

streams on a channel. Statistical multiplexing is a pro-

cess by which the instantaneous encoding rates of multi-

ple program streams are adjusted by VBR techniques to

maintain either (near) constant quality or optimize the

aggregate bit rate in the channel.

Network Transmission Issues

In networks with quality of service guarantees such as the

Internet, the varying availability of bandwidth leads to

network congestion and packet jitter. Approaches for ad-

dressing these issues include buffering the video stream

and employing techniques that yield multiple bit rates for

same stream. Hunt (2001) filed a patent application that

indicates a congestion management scheme—intended

primarily for mobile videophone users—in which graph-

ics and text may be displayed when it is necessary to

buffer the video stream so that the user perceives the video

streaming process to be in real time.

DIGITAL VIDEO COMPRESSION

STANDARDS

MPEG-1, -2, -4 Visual Codecs

Introduction

Koenen (2001) pointed out that the Moving Pic-

ture Expert Group (MPEG)—whose formal name is

ISO/IEC JTC1/SC29/WG11—has developed the widely

used MPEG-1, -2 and -4 codecs to facilitate and stan-

dardize infrastructure for the interoperability of multi-

media. MPEG-1 is still widely used on the Web, targeted

at bit rates 64 kbs –1.5 Mbps, MPEG-2 is the basis

for current digital television and Digital Versatile Disk

(DVD) standards, targeted at 1.5 Mbps –7 Mbps or higher,

and MPEG-4 is an extensible toolbox of algorithms that

addresses the coding of audiovisual objects throughout

an extended bit range from about 5 bps to more than

1 Gbps. The standards address video compression, to-

gether with audio and systems elements. The MPEG Web

site (http://mpeg.telecomitalialab.com) provides much de-

tailed information about the MPEG standards including

an overview.

Intra-, Predictive, and Bidirectional Frames

Table 2 shows the relationship among intra (I), predic-

tive (P), and bidirectional (B) frames referred to in the

MPEG standards and introduced in the MPEG-1 standard

(ISO/IEC 11172–2:1993). I frames are coded using infor-

mation only from that frame. P frames are coded using

forward prediction from previous I or P frames. B frames

are encoded from previous and succeeding frames using I

or P frames. This, in turn, implies that the encoding order

may be markedly different from the display order. Since B

frames require both past and future frames to be decoded

before they can be decoded and displayed, the presence

of B frames adds considerably to the encoder delay. The

MPEG-2 simple profile at main level requires only I and

P frames, thus reducing delay.

Sequence, Group of Pictures, Slice, Macroblock,

Block, and Pixel

A group of pictures (GOP) contains I, and potentially P,

or B frames. During encoding the user selects the du-

ration of the GOP as the number frames between suc-

cessive I frames (N) together with the distance between

Table 2Display and Transmission of a Group of 18 NTSC MPEG-2 Intra (I)

Display I B P B P B P B P B P B P B P B P B

Frame 1 2 345678 9101112131415161718

Transmit B I B P B P B P B P B P B P B P B P

Frame 2 1 43658710 91211141316151817

Note: Predicted (P) and bidirectional (B) pictures. The P pictures are predicted from the previous I picture and the B pictures

are interpolated from both the previous and next I or P pictures. The decoder must buffer P frames and compute B frames.

The maximum length of a group of pictures (GOP) is 18 pictures on the NTSC system and 15 pictures on the PAL system.