
1-8
Introduction
Final Rev F
Copyright 1996 by LSI Logic Corporation. All rights reserved.
current macroblock based on past and future reference pictures that the
encoder has already stored. Coefcient data is then inverse quantized
and operated on by an inverse DCT process that changes data from the
frequency domain to the time and space domain.
After the decoder processes all of the macroblocks, the picture recon-
struction is complete. If the picture just reconstructed is a reference pic-
ture (I-picture or P-picture), it replaces the oldest stored reference picture
and is used as the new reference for subsequent pictures. The pictures
may need to be reordered before they are displayed, in accordance with
the display order instead of the coding order. After the pictures are reor-
dered, they may be displayed on an appropriate output device.
1.2
Audio
Compression
and
Decompression
Concepts
Given an
elementary stream of data (for audio data, this is called an
audio stream), an MPEG encoder rst digitally compresses and codes
the data. The MPEG algorithm offers a choice of levels of complexity and
performance for this process.
To prepare a stream of compressed audio data for transmission, it is for-
matted into
audio frames. Each audio frame contains audio data, error-
correction data, and optional user-dened
ancillary data. The audio
frames are then sent in
packets grouped within packs in an ISO MPEG
System Stream.
The packs in system streams may contain a mix of audio packets and
video packets for one or more channels. Packs may contain packets from
separate elementary streams. Thus, MPEG can easily support multiple
channels of program material, and a decoder given access to a system
stream may access large numbers of channels.
1.2.1
MPEG Audio
Encoding
MPEG audio encoding is intended to efciently represent a digitized
audio stream by removing redundant information. Because different
applications have different performance goals, MPEG uses different
encoding techniques. These techniques, called
Layers, provide a differ-
ent trade-off between compression and signal quality. The MPEG algo-
rithm uses the two following processes for removing redundant audio
information:
Coding and quantization
Psychoacoustic modelling