2.7.4 Bitstream Structure

2.7.4.1 Hierarchical Coding Schemes

Hierarchical coding such as that used in MPEG compressed video, allows a 'quality hierarchy' to be implemented within a bitstream, where the lowest levels of the hierarchy contain the minimum information for intelligibility [Crowcroft, 1995]. Higher layers are used to refine the base image. This mechanism is ideal for transmission over packet-switched networks, where packets can be sent with information within the header as to the importance of the information they carry, allowing less essential packets to be dropped if the capacity of the channel or network is exceeded, and switching delay problems to be minimised by giving a higher priority to routing packets of the lowest level. Packets would carry data from one layer only, marked with their importance to the end user. Priority information is not uncommon; it already exists in protocols such as IP (Internet Protocol). In MPEG schemes, there are three bitstreams, each with their own structure and substructure. These are the system, video and audio streams and are detailed below. Dedicated system, video and audio decoders are generally used in MPEG-decoding hardware and each stream is decoded acording to the relevant part of the MPEG standard, i.e. part 1 for systems, 2 for video and 3 for audio (see figure 15).

Figure 15. Structure of an MPEG stream decoder.

2.7.4.2 System Stream Structure

The system stream is a combination of an MPEG video and audio stream, with added information. This added information forms the first of two layers with the system stream - the system layer. The second layer, known as the compression layer contains the audio and video bitstreams themselves. The main purpose of the system stream is to provide information to the decoder in order that the bitstreams contained within it can be decoded and synchronised correctly. In an MPEG-decoding system, the system decoder has the task of demultiplexing the audio and video bitstreams and routing them to the appropriate decoder. It also extracts timing information, such as timestamps and clock references for the purpose of synchronising the audio and video outputs.

Last page