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

DIGITALVIDEOCOMPRESSIONSTANDARDS 545

nearest I or P frames (M). Prediction errors are propa-

gated throughout the GOP and accumulate until the next

I-frame is reached. IfNis large, the accumulated error

may become unacceptable. IfMis 1, the encoder will use

only I and P frames.

A frame may be divided into slices, which contains

rows of macroblocks, from a single macroblock row to

a full frame. A macroblock contains the luminance infor-

mationYfor a 2×2 block area together with the subsam-

pled chrominance informationCB,CR, which, for 4:2:0

subsampled video sequences, may cover as small as one

block. Each block, in turn is defined to contain 8× 8

pixels over which discrete cosine transform coding is

performed (see Data Compression). MPEG-2 Test Model 5

rate control depends on adjusting the quantization of the

transform coefficients resulting from all the blocks of a

macroblock slice in a feedback loop.

MPEG-4 System Decoder

In the same way that MPEG-1 and -2 standardize only

the bit-stream syntax and decoder algorithms, MPEG-4

standardizes a system decoder model. The Koenen (2002)

overview of the MPEG-4 standard showed the concept

of encoding scenes containing visual objects and sprites.

The discrete cosine transform, quantization, inverse dis-

crete cosine transform, inverse quantization (see Data

Compression), shape coding, motion estimation, motion

prediction, and motion texture coding all combine to

optimize the visual layer of the video compression system

for specific content.

Error Resilience

Video compression typically uses interframe compres-

sion techniques to optimize the bit rate. Furthermore,

the channel coder receives variable length codes as input.

Any transmission errors, to which channels without qual-

ity of service guarantees (such as the wireless Internet)

are especially prone, may cause loss of synchronization

and the inability to reconstruct certain frames. Error re-

silience techniques such as those defined in the MPEG-4

and H.263 standards serve to reduce this risk. Resynchro-

nization markers are inserted periodically at the start of

a video packet, based on a predetermined threshold num-

ber of encoded bits and reversible variable length codes

allow some of the data between resynchronization mark-

ers to be recovered if the data have been corrupted by

errors.

Error Concealment

Koenen (2002) pointed out that a data partitioning ap-

proach can improve on the simple expedient of copying

blocks from a previous frame when errors have occurred.

A second synchronization marker is inserted into the bit

stream between the motion and the texture information.

When errors occur, the texture information is discarded

and the motion information used to compensate the pre-

vious decoded video packet. In real-time situations, where

there is a backchannel from the decoder to the encoder,

dynamic resolution conversion may be used to stabilize

the transmission buffering delay.

Content-Based Compression

The MPEG-4 standard supports content-based coding,

random-access to content objects, and extended manip-

ulation of content objects.

Shape and Texture Coding

The shape-adaptive discrete cosine transform based on

predefined orthonormal sets of one-dimensional discrete

cosine transform functions (see Kaup & Panis, 1997) can

be used to encode visual objects of arbitrary shape (not

just rectangles) together with texture.

Sprite Coding

MPEG-4 supports syntax for the efficient coding of static

and dynamically generated sprites, which are still images

representing backgrounds visible throughout a scene of a

video sequence.

Object Coding

MPEG-4 supports the coding of audiovisual objects.

Figure 3 depicts an audiovisual scene containing scrolling

text, audio, background sprite, arbitrarily shaped video,

and graphics.

General Scalability

The MPEG-4 Standard supports the scalability of visual

objects through the following profiles described in Table 3.

MPEG-7 Visual and MPEG-21 Standards

The MPEG-7 multimedia content description interface

(Martinez, 2001) supports query-by-content via descrip-

tors expressed in the extensible markup language. In the

visual case, the standard intends to convey basic and so-

phisticated information about the color, texture, shape,

motion, localization of visual objects, and the recognition

of faces.

The stated vision for the MPEG-21 multimedia frame-

work (Bormans & Hill, 2000) is “to enable transparent

and augmented use of multimedia resources across a wide

range of networks and devices.” The standard is intended

to support interoperable content representation and in-

tellectual property rights management in a “scalable and

error resilient way. The content representation of the me-

dia resources shall be synchronisable and multiplexed and

allow interaction.”

MPEG-4 includes hooks for an open intellectual

property management and protection scheme. A more in-

teroperable solution is planned for development in the

MPEG-21 standard. Bormans and Hill (2000) further

described specific interactions and showed how the

MPEG-7 standard supports transactions that produce and

consume digital data items.

ITU-T Visual Codecs

The ITU visual codecs H.261 and H.263, Video Codec for

Audio Visual Services at px64 kbit/s and Video Coding

for Low Bit-Rate Communications, are primarily used

for video conferencing applications, in which data rate

and end-to-end delay are important for lip synchroniza-

tion, a situation not encountered in broadcast applica-

tions. However, MPEG-4 contains many concepts derived