INTERNATIONAL ORGANISATION FOR
STANDARDISATION
ORGANISATION INTERNATIONALE DE NORMALISATION
ISO/IEC JTC1/SC29/WG11
CODING OF MOVING PICTURES AND AUDIO
ISO/IEC JTC1/SC29/WG11 N3156
December 1999/Maui
Source: WG11 (MPEG)
Status: Final
Title: MPEG-4
Overview - (Maui Version)
Editor: Rob Koenen
Overview of the MPEG-4 Standard
MPEG-4 is an ISO/IEC standard developed by MPEG (Moving Picture Experts Group), the committee that also developed the Emmy Award winning standards known as MPEG-1 and MPEG-2. These standards made interactive video on CD-ROM and Digital Television possible. MPEG-4 is the result of another international effort involving hundreds of researchers and engineers from all over the world. MPEG-4, whose formal ISO/IEC designation is ISO/IEC 14496, was finalized in October 1998 and became an International Standard in the first months of 1999. The fully backward compatible extensions under the title of MPEG-4 Version 2 were frozen at the end of 1999, to acquire the formal International Standard Status early 2000. Some work, on extensions in specific domains, is still in progress.
MPEG-4 builds on the proven success of three fields:
MPEG-4 provides the standardized technological elements enabling the integration of the production, distribution and content access paradigms of the three fields.
More information about MPEG-4 can be found at MPEG’s home page (case sensitive): . This web page contains links to a wealth of information about MPEG, including much about MPEG-4, many publicly available documents, several lists of ‘Frequently Asked Questions’ and links to other MPEG-4 web pages.
This document gives an overview of the MPEG-4 standard, explaining which pieces of technology it includes and what sort of applications are supported by this technology.
The MPEG-4 standard provides a set of technologies to satisfy the needs of authors, service providers and end users alike.
For all parties involved, MPEG seeks to avoid a multitude of proprietary, non-interworking formats and players.
MPEG-4 achieves these goals by providing standardized ways to:
The following sections illustrate the MPEG-4 functionalities described above, using the audiovisual scene depicted in Figure 1.
MPEG-4 audiovisual scenes are composed of several media objects, organized in a hierarchical fashion. At the leaves of the hierarchy, we find primitive media objects, such as:
MPEG-4 standardizes a number of such primitive media objects, capable of representing both natural and synthetic content types, which can be either 2- or 3-dimensional. In addition to the media objects mentioned above and shown in Figure 1, MPEG-4 defines the coded representation of objects such as:
A media object in its coded form consists of descriptive elements that allow handling the object in an audiovisual scene as well as of associated streaming data, if needed. It is important to note that in its coded form, each media object can be represented independent of its surroundings or background.
The coded representation of media objects is as efficient as possible while taking into account the desired functionalities. Examples of such functionalities are error robustness, easy extraction and editing of an object, or having an object available in a scaleable form.
Figure 1 explains the way in which an audiovisual scene in MPEG-4 is described as composed of individual objects. The figure contains compound media objects that group primitive media objects together. Primitive media objects correspond to leaves in the descriptive tree while compound media objects encompass entire sub-trees. As an example: the visual object corresponding to the talking person and the corresponding voice are tied together to form a new compound media object, containing both the aural and visual components of that talking person.
Such grouping allows authors to construct complex scenes, and enables consumers to manipulate meaningful (sets of) objects.
More generally, MPEG-4 provides a standardized way to describe a scene, allowing for example to:
The scene description builds on several concepts from the Virtual Reality Modeling language (VRML) in terms of both its structure and the functionality of object composition nodes and extends it to fully enable the aforementioned features.
Figure 1 - an example of an MPEG-4 Scene
Media objects may need streaming data, which is conveyed in one or more elementary streams. An object descriptor identifies all streams associated to one media object. This allows handling hierarchically encoded data as well as the association of meta-information about the content (called ‘object content information’) and the intellectual property rights associated with it.
Each stream itself is characterized by a set of descriptors for configuration information, e.g., to determine the required decoder resources and the precision of encoded timing information. Furthermore the descriptors may carry hints to the Quality of Service (QoS) it requests for transmission (e.g., maximum bit rate, bit error rate, priority, etc.)
Synchronization of elementary streams is achieved through time stamping of individual access units within elementary streams. The synchronization layer manages the identification of such access units and the time stamping. Independent of the media type, this layer allows identification of the type of access unit (e.g., video or audio frames, scene description commands) in elementary streams, recovery of the media object’s or scene description’s time base, and it enables synchronization among them. The syntax of this layer is configurable in a large number of ways, allowing use in a broad spectrum of systems.
The synchronized delivery of streaming information from source to destination, exploiting different QoS as available from the network, is specified in terms of the aforementioned synchronization layer and a delivery layer containing a two-layer multiplexer, as depicted in Figure 2.
The first multiplexing layer is managed according to the DMIF specification, part 6 of the MPEG-4 standard. (DMIF stands for Delivery Multimedia Integration Framework) This multiplex may be embodied by the MPEG-defined FlexMux tool, which allows grouping of Elementary Streams (ESs) with a low multiplexing overhead. Multiplexing at this layer may be used, for example, to group ES with similar QoS requirements, reduce the number of network connections or the end to end delay.
The "TransMux" (Transport Multiplexing) layer in Figure 2 models the layer that offers transport services matching the requested QoS. Only the interface to this layer is specified by MPEG-4 while the concrete mapping of the data packets and control signaling must be done in collaboration with the bodies that have jurisdiction over the respective transport protocol. Any suitable existing transport protocol stack such as (RTP)/UDP/IP, (AAL5)/ATM, or MPEG-2’s Transport Stream over a suitable link layer may become a specific TransMux instance. The choice is left to the end user/service provider, and allows MPEG-4 to be used in a wide variety of operation environments.
Figure 2 - The MPEG-4 System Layer Model
Use of the FlexMux multiplexing tool is optional and, as shown in Figure 2, this layer may be empty if the underlying TransMux instance provides all the required functionality. The synchronization layer, however, is always present.
With regard to Figure 2, it is possible to:
Parts of the control functionalities are available only in conjunction with a transport control entity like the DMIF framework.
In general, the user observes a scene that is composed following the design of the scene’s author. Depending on the degree of freedom allowed by the author, however, the user has the possibility to interact with the scene. Operations a user may be allowed to perform include:
More complex kinds of behavior can also be triggered, e.g. a virtual phone rings, the user answers and a communication link is established.
It is important to have the possibility to identify intellectual property in MPEG-4 media objects. Therefore, MPEG has worked with representatives of different creative industries in the definition of syntax and tools to support this. A full elaboration of the requirements for the identification of intellectual property can be found in ‘Management and Protection of Intellectual Property in MPEG-4, which is publicly available from the MPEG home page.
MPEG-4 incorporates identification the intellectual property by storing unique identifiers, that are issued by international numbering systems (e.g. ISAN, ISRC, etc. [ ISAN: International Audio-Visual Number, ISRC: International Standard Recording Code] ). These numbers can be applied to identify a current rights holder of a media object. Since not all content is identified by such a number, MPEG-4 Version 1 offers the possibility to identify intellectual property by a key-value pair (e.g.:»composer«/»John Smith«). Also, MPEG-4 offers a standardized interface that is integrated tightly into the Systems layer to people who want to use systems that control access to intellectual property. With this interface, proprietary control systems can be easily amalgamated with the standardized part of the decoder.
This Section contains a detailed overview of the MPEG-4 Standard. It first described the entire system, and then describes all parts of the system in subsections.
Figure 3 shows how streams coming from the network (or a storage device), as TransMux Streams, are demultiplexed into FlexMux Streams and passed to appropriate FlexMux demultiplexers that retrieve Elementary Streams. How this works is described in Section 2.2. The Elementary Streams (ESs) are parsed and passed to the appropriate decoders. Decoding recovers the data in an AV object from its encoded form and performs the necessary operations to reconstruct the original AV object ready for rendering on the appropriate device. Audio and visual objects are represented in their coded form, which is described in sections 2.4 and 2.5. The reconstructed AV object is made available to the composition layer for potential use during scene rendering. Decoded AVOs, along with scene description information, are used to compose the scene as described by the author. Scene description and Composition are explained in Section 2.6. The user can, to the extent allowed by the author, interact with the scene which is eventually rendered and presented. Section 2.7 describes this interaction. Finally, Sections 2.8 and 2.9 discuss the ‘Intellectual Property Management and Protection’ and Object Content Information respectively.
Figure 3- Major components of an MPEG-4 terminal (receiver side)
DMIF (Delivery Multimedia Integration Framework) is a session protocol for the management of multimedia streaming over generic delivery technologies. In principle it is similar to FTP. The only (essential!) difference is that FTP returns data, DMIF returns pointers to where to get (streamed) data
When FTP is run, the very first action it performs is the setup of a session with the remote side. Later, files are selected and FTP sends a request to download them, the FTP peer will return the files in a separate connection.
Similarly, when DMIF is run, the very first action it performs is the setup of a session with the remote side. Later, streams are selected and DMIF sends a request to stream them, the DMIF peer will return the pointers to the connections where the streams will be streamed, and then also establishes the connection themselves.
Compared to FTP, DMIF is both a framework and a protocol. The functionality provided by DMIF is expressed by an interface called DMIF-Application Interface (DAI), and translated into protocol messages. These protocol messages may differ based on the network on which they operate.
The Quality of Service is also considered in the DMIF design, and the DAI allows the DMIF user to specify the requirements for the desired stream. It is then up to the DMIF implementation to make sure that the requirements are fulfilled. The DMIF specification provides hints on how to perform such tasks on a few network types, such as the Internet.
The DAI is also used for accessing broadcast material and local files, this means that a single, uniform interface is defined to access multimedia contents on a multitude of delivery technologies.
As a consequence, it is appropriate to state that the integration framework of DMIF covers three major technologies, interactive network technology, broadcast technology and the disk technology; this is shown in the Figure 4 below.
Figure 4 - DMIF addresses the delivery integration of three major technologies
The DMIF architecture is such that applications that rely on DMIF for communication do not have to be concerned with the underlying communication method. The implementation of DMIF takes care of the delivery technology details presenting a simple interface to the application.
Figure 4 represents the above concept. An application accesses data through the DMIF-Application Interface, irrespective of whether such data comes from a broadcast source, from local storage or from a remote server. In all scenarios the Local Application only interacts through a uniform interface (DAI). Different DMIF instances will then translate the Local Application requests into specific messages to be delivered to the Remote Application, taking care of the peculiarities of the involved delivery technology. Similarly, data entering the terminal (from remote servers, broadcast networks or local files) is uniformly delivered to the Local Application through the DAI.
Different, specialized DMIF instances are indirectly invoked by the Application to manage the various specific delivery technologies, this is however transparent to the Application, that only interacts with a single "DMIF filter". This filter is in charge of directing the particular DAI primitive to the right instance. DMIF does not specify this mechanism, just assumes it is implemented. This is further emphasized by the shaded boxes in the figure, whose aim is to clarify what are the borders of a DMIF implementation, while the DMIF communication architecture defines a number of modules, actual DMIF implementations only need to preserve their appearance at those borders.
Conceptually, a "real" remote application accessed through a network e.g., IP- or ATM-based, is no different than an emulated remote producer application getting content from a broadcast source or from a disk. In the former case, however, the messages exchanged between the two entities have to be normatively defined to ensure interoperability (these are the DMIF Signaling messages). In the latter case, on the other hand, the interfaces between the two DMIF peers and the emulated Remote Application are internal to a single implementation and need not be considered in this specification. Note that for the broadcast and local storage scenarios, the figure shows a chain of "Local DMIF", "Remote DMIF (emulated)" and "Remote Application (emulated)". This chain only represents a conceptual model and need not be reflected in actual implementations (it is shown in the figure totally internal to a shaded box).
Figure 5 - DMIF communication architecture
When considering the Broadcast and Local Storage scenarios, it is assumed that the (emulated) Remote Application has knowledge on how the data is delivered/stored. This implies knowledge of the kind of application it is dealing with. In the case of MPEG-4, this actually means knowledge of concepts like Elementary Stream ID, First Object Descriptor, ServiceName. Thus, while the DMIF Layer is conceptually unaware of the application it is providing support to, in the particular case of DMIF instances for Broadcast and Local Storage this assumption is not completely true due to the presence of the (emulated) Remote Application (which, from the Local Application perspective, is still part of the DMIF Layer).
It is worth noting that since the (emulated) Remote Application has knowledge on how the data is delivered/stored, the specification of how data is delivered/stored is crucial for such a DMIF implementation, which is thus "MPEG-4 systems aware".
When considering the Remote Interactive scenario instead, the DMIF Layer is totally application-unaware. An additional interface -the DMIF-Network Interface (DNI)- is introduced to emphasize what kind of information DMIF peers need to exchange; an additional module ("Signaling mapping" in the figure) takes care of mapping the DNI primitives into signaling messages used on the specific Network. Note that DNI primitives are only specified for information purposes, and a DNI interface need not be present in an actual implementation, Figure 5 also clearly represents the DNI as internal to the shaded box. Instead, the syntax of the messages flowing in the Network is fully specified for each specific network supported.
DMIF allows the concurrent presence of one or more DMIF instances, each one targeted for a particular delivery technology, in order to support in the same terminal multiple delivery technologies and even multiple scenarios (broadcast, local storage, remote interactive). Multiple delivery technologies may be activated by the same application, that could therefore seamlessly manage data sent by broadcast networks, local file systems and remote interactive peers
When an application requests the activation of a service, it uses the Service primitives of the DAI, and creates a service session; the DMIF implementation then contacts its corresponding peer (that conceptually can be either a remote peer, or a local emulated peer) and creates a network session with it. Network sessions have network-wide significance, service sessions have instead local meaning. The association between them is maintained by the DMIF Layer. In the case of Broadcast and Local Storage scenarios, the way the network session is created and then managed is out of the scope of this specification. In the case of a remote interactive scenario instead, DMIF uses the native signalling mechanism for that network to create and then manage the network session e.g., ATM signalling. The application peers then use this session to create connections which are used to transport application data e.g., MPEG-4 Elementary Streams.
When an application needs a Channel, it uses the Channel primitives of the DAI, DMIF translates these requests into connection requests which are specific to the particular network implementation. In the case of Broadcast and Local Storage scenarios, the way the connections are created and then managed is out of the scope of this specification. In the case of a networked scenario instead, DMIF uses the native signalling mechanism for that network to create those connections. The application then uses these connections to deliver the service.
Figure 6 provides a high level view of a service activation and of the beginning of data exchange; the high level walk-through consists of the following steps:
The Originating Application request the activation of a service to its local DMIF Layer -- a communication path between the Originating Application and its local DMIF peer is established in the control plane (1)
The Originating DMIF peer establishes a network session with the Target DMIF peer -- a communication path between the Originating DMIF peer and the Target DMIF Peer is established in the control plane (2)
The Target DMIF peer identifies the Target Application and forwards the service activation request -- a communication path between the Target DMIF peer and the Target Application is established in the control plane (3)
The peer Applications create channels (requests flowing through communication paths 1, 2 and 3). The resulting channels in the user plane (4) will carry the actual data exchanged by the Applications.
DMIF is involved in all four steps above.
Figure 6 — DMIF Computational Model
The DMIF Layer automatically determines whether a particular service is supposed to be provided by a remote server on a particular network e.g., IP based, or ATM based, by a broadcast network, or resides in a local storage device: the selection is based on the peer address information provided by the Application as part of a URL passed to the DAI.
Individual Elementary Streams have to be retrieved on the delivery layer from incoming data from some network connection or a storage device. Each network connection or file is homogeneously considered a TransMux Channel in the MPEG-4 system model. The demultiplexing is partially or completely done by layers outside the scope of MPEG-4, depending on the application. The only multiplexing tool defined by MPEG-4 is the FlexMux tool that may optionally be used for low delay, low overhead multiplexing and for saving network connection resources.
For the purpose of integrating MPEG-4 in system environments, the DMIF Application Interface is the reference point at which elementary streams can be accessed as sync layer-packetized streams. The DMIF Network Interface specifies how either SL(Sync Layer)-packetized streams —no FlexMux used— or FlexMux Streams are to be retrieved from the TransMux Layer. This is the interface to the transport functionalities not defined by MPEG. The data part of the interfaces is considered here, while the control part is dealt with by DMIF.
In the same way that MPEG-1 and MPEG-2 describe the behavior of an idealized decoding device along with the bitstream syntax and semantics, MPEG-4 defines a System Decoder Model. This allows the precise definition of the terminal’s operation without making unnecessary assumptions about implementation details. This is essential in order to give implementers the freedom to design real MPEG-4 terminals and decoding devices in a variety of ways. These devices range from television receivers, which have no ability to communicate with the sender, to computers that are fully enabled with bi-directional communication. Some devices will receive MPEG-4 streams over isochronous networks, while others will use non-isochronous means (e.g., the Internet) to exchange MPEG-4 information. The System Decoder Model provides a common model on which all implementations of MPEG-4 terminals can be based.
The specification of a buffer and timing model is essential to encoding devices which may not know ahead of time what the terminal device is or how it will receive the encoded stream. Though the MPEG-4 specification will enable the encoding device to inform the decoding device of resource requirements, it may not be possible, as indicated earlier, for that device to respond to the sender. It is also possible that an MPEG-4 session is received simultaneously by widely different devices; it will, however, be properly rendered according to the capability of each device.
Demultiplexing occurs on the delivery layer that is modeled as consisting of a TransMux layer and a DMIF layer. The retrieval of incoming data streams from network connections or storage media consists of two tasks. First, the channels must be located and opened. This requires a transport control entity that manages, among others, the tables that associate transport channels to specific elementary streams. Stream map tables link each stream to a ChannelAssociationTag that serves as a handle to the channel that carries this stream. Resolving ChannelAssociationTags to the actual transport channel as well as the management of the sessions and channels is addressed by the DMIF part of the MPEG-4 standard.
Second, the incoming streams must be properly demultiplexed to recover SL-packetized streams from downstream channels (incoming at the receiving terminal) to be passed on to the synchronization layer. In interactive applications, a corresponding multiplexing stage will multiplex upstream data in upstream channels (outgoing from the receiving terminal).
The generic term ‘TransMux Layer’ is used to abstract any underlying multiplex functionality – existing or future – that is suitable to transport MPEG-4 data streams. Note that this layer is not defined in the context of MPEG-4. Examples are MPEG-2 Transport Stream, H.223, ATM AAL 2, IP/UDP. The TransMux Layer is assumed to provide protection and multiplexing functionality, indicating that this layer is responsible for offering a specific QoS. Protection functionality includes error protection and error detection tools suitable for the given network or storage medium.
In any concrete application scenario one or more specific TransMux Instances will be used. Each TransMux demultiplexer gives access to TransMux Channels. The requirements on the data interface to access a TransMux Channel are the same for all TransMux Instances. They include the need for reliable error detection, delivery, if possible, of erroneous data with a suitable error indication and framing of the payload, which may consist of either SL-packetized streams or FlexMux streams. These requirements are summarized in an informative way in the TransMux Interface, in the Systems part of the MPEG-4 Standard. An adaptation of SL-packetized streams must be specified to each transport protocol stack of interest according to these requirements and in conjunction with the standardization body that has the proper jurisdiction. This is happening for RTP and mobile channels at the moment.
The FlexMux tool is specified by MPEG to optionally provide a flexible, low overhead, low delay method for interleaving data whenever this is not sufficiently supported by the underlying protocol stack. It is especially useful when the packet size or overhead of the underlying TransMux instance is large, so that a waste of bandwidth or number of network connections would result otherwise. The FlexMux tool is not itself robust to errors and can either be used on TransMux Channels with a high QoS or to bundle Elementary Streams that are equally error tolerant. The FlexMux requires reliable error detection and sufficient framing of FlexMux packets (for random access and error recovery) from the underlying layer. These requirements are also reflected in the data primitives of the DMIF Application Interface, which defines the data access to individual transport channels. The FlexMux demultiplexer retrieves SL-packetized streams from FlexMux Streams.
Figure 7 - Buffer architecture of the System Decoder Model
The sync layer has a minimum set of tools for consistency checking, padding, to convey time base information and to carry time stamped access units of an elementary stream. Each packet consists of one access unit or a fragment of an access unit. These time stamped access units form the only semantic structure of elementary streams that is visible on this layer. Time stamps are used to convey the nominal decoding and composition time for an access unit. The sync layer requires reliable error detection and framing of each individual packet from the underlying layer, which can be accomplished, e.g., by using the FlexMux. How data can be accessed by the compression layer is summarized in the informative Elementary Stream Interface, which can be found in the Systems part of the MPEG-4 Standard. The sync layer retrieves elementary streams from SL-packetized streams.
To be able to relate elementary streams to media objects within a scene, object descriptors are used. Object Descriptors convey information about the number and properties of elementary streams that are associated to particular media objects. Object descriptors are themselves conveyed in one or more elementary streams, since it is possible to add and discard streams (and objects) during the course of an MPEG-4 session. Such updates are time stamped in order to guarantee synchronization. The object descriptor streams can be considered as a description of the streaming resources for a presentation. Similarly, the scene description is also conveyed as an elementary stream, allowing to modify the spatio-temporal layout of the presentation over time.
To predict how the decoder will behave when it decodes the various elementary data streams that form an MPEG-4 session, the Systems Decoder Model enables the encoder to specify and monitor the minimum buffer resources that are needed to decode a session. The required buffer resources are conveyed to the decoder within object descriptors during the setup of the MPEG-4 session, so that the decoder can decide whether it is capable of handling this session.
By managing the finite amount of buffer space the model allows a sender, for example, to transfer non real-time data ahead of time, if sufficient space is available at the receiver side to store it. The pre-stored data can then be accessed when needed, allowing at that time real-time information to use a larger amount of the channel’s capacity if so desired.
For real-time operation, a timing model is assumed in which the end-to-end delay from the signal output from an encoder to the signal input to a decoder is constant. Furthermore, the transmitted data streams must contain implicit or explicit timing information. There are two types of timing information. The first is used to convey the speed of the encoder clock, or time base, to the decoder. The second, consisting of time stamps attached to portions of the encoded AV data, contains the desired decoding time for access units or composition and expiration time for composition units. This information is conveyed in SL-packet headers generated in the sync layer. With this timing information, the inter-picture interval and audio sample rate can be adjusted at the decoder to match the encoder’s inter-picture interval and audio sample rate for synchronized operation.
Different media objects may have been encoded by encoders with different time bases, with the accompanying slightly different speed. It is always possible to map these time bases to the time base of the receiving terminal. In this case, however, no real implementation of a receiving terminal can avoid the occasional repetition or drop of AV data, due to temporal aliasing (the relative reduction or extension of their time scale).
Although systems operation without any timing information is allowed, defining a buffering model is not possible for this case.
MPEG-4 defines a syntactic description language to describe the exact binary syntax for bitstreams carrying media objects and for bitstreams with scene description information. This is a departure from MPEG’s past approach of utilizing pseudo C. This language is an extension of C++, and is used to describe the syntactic representation of objects and the overall media object class definitions and scene description information in an integrated way. This provides a consistent and uniform way of describing the syntax in a very precise form, while at the same time simplifying bitstream compliance testing. Software tools can be used to process the syntactic description and generate the necessary code for programs that perform validation.
MPEG-4 coding of audio objects provides tools for both representing natural sounds (such as speech and music) and for synthesizing sounds based on structured descriptions. The representation for synthesized sound can be derived from text data or so-called instrument descriptions and by coding parameters to provide effects, such as reverberation and spatialization. The representations provide compression and other functionalities, such as scalability and effects processing.
The MPEG-4 Audio coding tools covering 6kbit/s to 24kbit/s have undergone verification testing for an AM digital audio broadcasting application in collaboration with the NADIB (Narrow Band Digital Broadcasting) consortium. With the intent of identifying a suitable digital audio broadcast format to provide improvements over the existing AM modulation services, several codec configurations involving the MPEG-4 CELP, TwinVQ, and AAC tools have been compared to a reference AM system. (see below for an explanation about these algorithms.) It was found that higher quality can be achieved in the same bandwidth with digital techniques and that scalable coder configurations offered performance superior to a simulcast alternative. Additional verification tests were carried out by MPEG, in which the tools for speech and general audio coding were compared to existing standards.
MPEG-4 standardizes natural audio coding at bitrates ranging from 2 kbit/s up to and above 64 kbit/s. When variable rate coding is allowed, coding at less than 2 kbit/s, such as an average bitrate of 1.2 kbit/s, is also supported. The presence of the MPEG-2 AAC standard within the MPEG-4 tool set provides for general compression of audio in the upper bitrate range. For these, the MPEG-4 standard defines the bitstream syntax and the decoding processes in terms of a set of tools. In order to achieve the highest audio quality within the full range of bitrates and at the same time provide the extra functionalities, speech coding techniques and general audio coding techniques are integrated in a common framework:
• Speech coding at bitrates between 2 and 24 kbit/s is supported by using Harmonic Vector eXcitation Coding (HVXC) for a recommended operating bitrate of 2 - 4 kbit/s, and Code Excited Linear Predictive (CELP) coding for an operating bitrate of 4 - 24 kbit/s. In addition, HVXC can operate down to an average of around 1.2 kbit/s in its variable bitrate mode. In CELP coding, two sampling rates, 8 and 16 kHz, are used to support narrowband and wideband speech, respectively. The following operating modes have been subject to verification testing: HVXC at 2 and 4 kbit/s, narrowband CELP at 6, 8.3, and 12 kbit/s, and wideband CELP at 18 kbit/s. In addition various of the scalable configurations have been verified.
• For general audio coding at bitrates at and above 6 kbit/s, transform coding techniques, namely TwinVQ and AAC, are applied. The audio signals in this region typically have sampling frequencies starting at 8 kHz.
To allow optimum coverage of the bitrates and to allow for bitrate and bandwidth scalability, a general framework has been defined. This is illustrated in Figure 8.
Figure 8 - General block diagram of MPEG-4 Audio
Starting with a coder operating at a low bitrate, by adding enhancements to a general audio coder, both the coding quality as well as the audio bandwidth can be improved.
Bitrate scalability, often also referred to as embedded coding, allows a bitstream to be parsed into a bitstream of lower bitrate that can still be decoded into a meaningful signal. The bitstream parsing can occur either during transmission or in the decoder. Bandwidth scalability is a particular case of bitrate scalability whereby part of a bitstream representing a part of the frequency spectrum can be discarded during transmission or decoding.
Encoder complexity scalability allows encoders of different complexity to generate valid and meaningful bitstreams. The decoder complexity scalability allows a given bitstream to be decoded by decoders of different levels of complexity. The audio quality, in general, is related to the complexity of the encoder and decoder used. Scalability works within some MPEG-4 tools, but can also be applied to a combination of techniques, e.g. with CELP as a base layer and AAC for the enhancement layer(s).
The MPEG-4 systems allows codecs according to existing (MPEG) standards, e.g. MPEG-2 AAC, to be used. Each of the MPEG-4 coders is designed to operate in a stand-alone mode with its own bitstream syntax. Additional functionalities are realized both within individual coders, and by means of additional tools around the coders. An example of such a functionality within an individual coder is speed or pitch change within HVXC.
MPEG-4 defines decoders for generating sound based on several kinds of ‘structured’ inputs. Text input is converted to speech in the Text-To-Speech (TTS) decoder, while more general sounds including music may be normatively synthesized. Synthetic music may be delivered at extremely low bitrates while still describing an exact sound signal.
Text To Speech. TTS coders bitrates range from 200 bit/s to 1.2 Kbit/s, which allows a text or a text with prosodic parameters (pitch contour, phoneme duration, and so on) as its inputs to generate intelligible synthetic speech. It supports the generation of parameters that can be used to allow synchronization to associated face animation, international languages for text and international symbols for phonemes. Additional markups are used to convey control information within texts, which is forwarded to other components in synchronization with the synthesized text. Note that MPEG-4 provides a standardized interface for the operation of a Text To Speech coder (TTSI = Text To Speech Interface), but not a normative TTS synthesizer itself.
Score Driven Synthesis.
The Structured Audio tools decode input data and produce output sounds. This decoding is driven by a special synthesis language called SAOL (Structured Audio Orchestra Language) standardized as a part of MPEG-4. This language is used to define an "orchestra" made up of "instruments" (downloaded in the bitstream, not fixed in the terminal) which create and process control data. An instrument is a small network of signal processing primitives that might emulate some specific sounds such as those of a natural acoustic instrument. The signal-processing network may be implemented in hardware or software and include both generation and processing of sounds and manipulation of pre-stored sounds.
MPEG-4 does not standardize "a single method" of synthesis, but rather a way to describe methods of synthesis. Any current or future sound-synthesis method can be described in SAOL, including wavetable, FM, additive, physical-modeling, and granular synthesis, as well as non-parametric hybrids of these methods.
Control of the synthesis is accomplished by downloading "scores" or "scripts" in the bitstream. A score is a time-sequenced set of commands that invokes various instruments at specific times to contribute their output to an overall music performance or generation of sound effects. The score description, downloaded in a language called SASL (Structured Audio Score Language), can be used to create new sounds, and also include additional control information for modifying existing sound. This allows the composer finer control over the final synthesized sound. For synthesis processes that do not require such fine control, the established MIDI protocol may also be used to control the orchestra.
Careful control in conjunction with customized instrument definition, allows the generation of sounds ranging from simple audio effects, such as footsteps or door closures, to the simulation of natural sounds such as rainfall or music played on conventional instruments to fully synthetic sounds for complex audio effects or futuristic music.
For terminals with less functionality, and for applications which do not require such sophisticated synthesis, a "wavetable bank format" is also standardized. Using this format, sound samples for use in wavetable synthesis may be downloaded, as well as simple processing, such as filters, reverbs, and chorus effects. In this case, the computational complexity of the required decoding process may be exactly determined from inspection of the bitstream, which is not possible when using SAOL.
Visual objects can be either of natural or of synthetic origin. First, the objects of natural origin are described. Then follow the synthetic ones.
The tools for representing natural video in the MPEG-4 visual standard provide standardized core technologies allowing efficient storage, transmission and manipulation of textures, images and video data for multimedia environments. These tools allow the decoding and representation of atomic units of image and video content, called "video objects" (VOs). An example of a VO could be a talking person (without background), which can then be composed with other AVOs (audio-visual objects) to create a scene. Conventional rectangular imagery is handled as a special case of such objects.
In order to achieve this broad goal rather than a solution for a narrow set of applications, functionalities common to several applications are clustered. Therefore, the visual part of the MPEG-4 standard provides solutions in the form of tools and algorithms for:
Synthetic objects form a subset of the larger class of computer graphics, as an initial focus the following visual synthetic objects will be described:
The ‘facial animation object’ can be used to render an animated face. The shape, texture and expressions of the face are controlled by Facial Definition Parameters (FDPs) and/or Facial Animation Parameters (FAPs). Upon construction, the face object contains a generic face with a neutral expression. This face can already be rendered. It can also immediately receive the animation parameters from the bitstream, which will produce animation of the face: expressions, speech etc. Meanwhile, definition parameters can be sent to change the appearance of the face from something generic to a particular face with its own shape and (optionally) texture. If so desired, a complete face model can be downloaded via the FDP set.
Face Animation in MPEG-4 Version 1 provides for highly efficient coding of animation parameters that can drive an unlimited range of face models. The models themselves are not normative, although (see above) there are normative tools to describe the appearance of the model. Frame-based and temporal-DCT coding of a large collection of FAPs can be used for accurate speech articulation. Viseme and expression parameters are used to code specific speech configurations of the lips and the mood of the speaker.
The Systems Binary Format for Scenes (BIFS, see section 2.6) provides features to support Face Animation when custom models and specialized interpretation of FAPs are needed:
These specialized node types in BIFS effectively provide for tailored face models including calibration of an established face model in a terminal or downloading of a fully custom model including its shape, texture, and color.
Body Animation, to be standardized in MPEG-4 Version 2, is being designed into the MPEG-4 fabric to work in a thoroughly integrated way with face/head animation. Decoding and scene description for Body Animation directly mirror technology already proven in Face Animation. Also here, Body Definition Parameters (BDPs) and Body Animation Parameters (BAPs) exist.
A 2-D mesh is a tessellation (or partition) of a 2-D planar region into polygonal patches. The vertices of the polygonal patches are referred to as the node points of the mesh. MPEG4 considers only triangular meshes where the patches are triangles. A 2-D dynamic mesh refers to 2-D mesh geometry and motion information of all mesh node points within a temporal segment of interest. Triangular meshes have long been used for efficient 3-D object shape (geometry) modeling and rendering in computer graphics. 2-D mesh modeling may be considered as projection of such 3-D triangular meshes onto the image plane. An example of a 2-D mesh is depicted in Figure 9.
Figure 9- 2-D mesh modeling of the "Bream"
video object.
By deforming the mesh, the fish can be animated very
efficiently, and be made to ‘swim’.
Also, a logo could be projected
onto the fish, and made to move in accordance with the fish
A dynamic mesh is a forward tracking mesh, where the node points of the initial mesh track image features forward in time by their respective motion vectors. The initial mesh may be regular, or can be adapted to the image content, which is called a content-based mesh. 2-D content-based mesh modeling then corresponds to non-uniform sampling of the motion field at a number of salient feature points (node points) along the contour and interior of a video object. Methods for selection and tracking of these node points are not subject to standardization.
In 2-D mesh based texture mapping, triangular patches in the current frame are deformed by the movements of the node points into triangular patches in the reference frame. The texture inside each patch in the reference frame is warped onto the current frame using a parametric mapping, defined as a function of the node point motion vectors. For triangular meshes, the affine mapping is a common choice. Its linear form implies texture mapping with low computational complexity. Affine mappings can model translation, rotation, scaling, reflection and shear, and preserve straight lines. The degrees of freedom given by the three motion vectors of the vertices of a triangle match with the six parameters of the affine mapping. This implies that the original 2-D motion field can be compactly represented by the motion of the node points, from which a continuous, piece-wise affine motion field can be reconstructed. At the same time, the mesh structure constrains movements of adjacent image patches. Therefore, meshes are well-suited to represent mildly deformable but spatially continuous motion fields.
2-D mesh modeling is attractive because 2-D meshes can be designed from a single view of an object without requiring range data, while maintaining several of the functionalities offered by 3-D mesh modeling. In summary, the 2-D object-based mesh representation is able to model the shape (polygonal approximation of the object contour) and motion of a VOP in a unified framework, which is also extensible to the 3-D object modeling when data to construct such models is available. In particular, the 2-D mesh representation of video objects enables the following functionalities:
A. Video Object Manipulation
B. Video Object Compression
C. Content-Based Video Indexing
The view-dependent scalability enables to stream texture maps, which are used in realistic virtual environments. It consists in taking into account the viewing position in the 3-D virtual world in order to transmit only the most visible information. Only a fraction of the information is then sent, depending on object geometry and viewpoint displacement. This fraction is computed both at the encoder and at the decoder side. This approach allows to reduce greatly the amount of transmitted information between a remote database and a user, given that a back-channel is available. This scalability can be applied both with Wavelet and DCT based encoders.
The MPEG-4 image and video coding algorithms give an efficient representation of visual objects of arbitrary shape, also supporting so-called content-based functionalities. They support most functionalities already provided by MPEG-1 and MPEG-2, including efficient compression of standard rectangular sized image sequences at varying levels of input formats, frame rates, pixel depth, bit-rates, and various levels of spatial, temporal and quality scalability.
A basic classification of the bit rates and functionalities currently provided by the MPEG-4 Visual standard for natural images and video is depicted in Figure 10 below, which clusters bit-rate levels versus sets of functionalities.
Figure 10 - Classification of the MPEG-4 Image and Video Coding Algorithms and Tools
At the bottom end a "VLBV Core" (VLBV: Very Low Bit-rate Video) provides algorithms and tools for applications operating at bit-rates typically between 5...64 kbits/s, supporting image sequences with low spatial resolution (typically up to CIF resolution) and low frame rates (typically up to 15 Hz). The basic applications specific functionalities supported by the VLBV Core include:
a) coding of conventional rectangular size image sequences with high coding efficiency and high error robustness/resilience, low latency and low complexity for real-time multimedia communications applications, and
b) "random access" and "fast forward" and "fast reverse" operations for VLB multimedia data-base storage and access applications.
The same basic functionalities outlined above are also supported at higher bit-rates with a higher range of spatial and temporal input parameters up to ITU-R Rec. 601 resolutions and larger - employing identical or similar algorithms and tools as the VLBV Core. The bit-rates envisioned range typically from 64 kbits/s up to 10 Mb/s and applications envisioned include multimedia broadcast or interactive retrieval of signals with a quality comparable to digital TV. For these applications at higher bit-rates, also interlaced can be represented by MPEG-4 coding tools.
Content-based functionalities support the separate encoding and decoding of content (i.e. physical objects in a scene, VOs). This MPEG-4 feature provides the most elementary mechanism for interactivity; flexible representation and manipulation with/of VO content of images or video in the compressed domain, without the need for further segmentation or transcoding at the receiver.
For the hybrid coding of natural as well as synthetic visual data (e.g. for virtual presence or virtual environments) the content-based coding functionality allows mixing a number of VO's from different sources with synthetic objects, such as a virtual backgrounds.
The extended MPEG-4 algorithms and tools for content-based functionalities can be seen as a superset of the VLBV core and high bit-rate tools - meaning that the tools provided by the VLBV and higher bitrate cores are complemented by additional elements.
As mentioned before, MPEG-4 Video supports conventional rectangular images and video as well as images and video of arbitrary shape. This concept is illustrated in Figure 11 below.
Figure 11 - the VLBV Core and the Generic MPEG-4 Coder
The coding of conventional images and video is similar to conventional MPEG-1/2 coding. It involves motion prediction/compensation followed by texture coding. For the content-based functionalities, where the image sequence input may be of arbitrary shape and location, this approach is extended by also coding shape and transparency information. Shape may be either represented by an 8 bit transparency component - which allows the description of transparency if one VO is composed with other objects - or by a binary mask.
The extended MPEG-4 content-based approach can be seen as a logical extension of the conventional MPEG-4 VLBV Core or high bit-rate tools towards input of arbitrary shape.
Figure 12 below outlines the basic approach of the MPEG-4 video algorithms to encode rectangular as well as arbitrarily shaped input image sequences.
Figure 12 - Basic block diagram of MPEG-4 Video Coder
The basic coding structure involves shape coding (for arbitrarily shaped VOs) and motion compensation as well as DCT-based texture coding (using standard 8x8 DCT or shape adaptive DCT).
The basic coding structure involves shape coding (for arbitrarily shaped VOs) and motion compensation as well as DCT-based texture coding (using standard 8x8 DCT or shape adaptive DCT).
An important advantage of the content-based coding approach MPEG-4 is that the compression efficiency can be significantly improved for some video sequences by using appropriate and dedicated object-based motion prediction "tools" for each object in a scene. A number of motion prediction techniques can be used to allow efficient coding and flexible presentation of the objects:
Figure 13 depicts the basic concept for coding an MPEG-4 video sequence using a sprite panorama image. It is assumed that the foreground object (tennis player, image top right) can be segmented from the background and that the sprite panorama image can be extracted from the sequence prior to coding. (A sprite panorama is a still image that describes as a static image the content of the background over all frames in the sequence). The large panorama sprite image is transmitted to the receiver only once as first frame of the sequence to describe the background – the sprite remains is stored in a sprite buffer. In each consecutive frame only the camera parameters relevant for the background are transmitted to the receiver. This allows the receiver to reconstruct the background image for each frame in the sequence based on the sprite. The moving foreground object is transmitted separately as an arbitrary-shape video object. The receiver composes both the foreground and background images to reconstruct each frame (bottom picture in figure below). For low delay applications it is possible to transmit the sprite in multiple smaller pieces over consecutive frames or to build up the sprite at the decoder progressively.
Figure 13 - Example of Sprite Coding of Video Sequence
Efficient Coding of visual textures and still images (e.g. to be mapped on animated meshes) is supported by the visual texture mode of the MPEG-4. This mode is based on a zerotree wavelet algorithm that provides very high coding efficiency over a very wide range of bitrates. Together with high compression efficiency, it also provides spatial and quality scalabilities (up to 11 levels of spatial scalability and continuous quality scalability) and also arbitrary-shaped object coding. The wavelet formulation provides for scalable bitstream coding in the form of an image resolution pyramid for progressive transmission and temporal enhancement of still images. The coded bitstream is also intended for downloading of the image resolution hierarchy into the terminal to be formatted as ‘MIPmap texture’ as used in 3-D rendering systems. This technology provides the resolution scalability to deal with a wide range of viewing conditions more typical of interactive applications and the mapping of imagery into 2-D and 3-D virtual worlds.
MPEG-4 supports the coding of images and video objects with spatial and temporal scalability, both with conventional rectangular as well as with arbitrary shape. Scalability refers to the ability to only decode a part of a bitstream and reconstruct images or image sequences with:
This functionality is desired for progressive coding of images and video sent over heterogeneous networks, as well as for applications where the receiver is not capable of displaying the full resolution or full quality images or video sequences. This could for instance happen when processing power or display resolution is limited.
For decoding of still images, the MPEG-4 standard will provide spatial scalability with up to 11 levels of granularity and also quality scalability up to the bit level. For video sequences a initial maximum of 3 levels of granularity will be supported, but work is ongoing to raise this level to 9.
MPEG-4 provides error robustness and resilience to allow accessing image or video information over a wide range of storage and transmission media. In particular, due to the rapid growth of mobile communications, it is extremely important that access is available to audio and video information via wireless networks. This implies a need for useful operation of audio and video compression algorithms in error-prone environments at low bit-rates (i.e., less than 64 kbit/s).
The error resilience tools developed for MPEG-4 can be divided into three major areas: resynchronization, data recovery, and error concealment. It should be noted that these categories are not unique to MPEG-4, but instead have been used by many researchers working in the area error resilience for video. It is, however, the tools contained in these categories that are of interest, and where MPEG-4 makes its contribution to the problem of error resilience.
Resynchronization tools attempt to enable resynchronization between the decoder and the bitstream after a residual error or errors have been detected. Generally, the data between the synchronization point prior to the error and the first point where synchronization is reestablished, is discarded. If the resynchronization approach is effective at localizing the amount of data discarded by the decoder, then the ability of other types of tools that recover data and/or conceal the effects of errors is greatly enhanced.
The resynchronization approach adopted by MPEG-4, referred to as a packet approach, is similar to the Group of Blocks (GOBs) structure utilized by the ITU-T standards of H.261 and H.263. In these standards a GOB is defined as one or more rows of macroblocks (MBs). At the start of a new GOB, information called a GOB header is placed within the bitstream. This header information contains a GOB start code, which is different from a picture start code, and allows the decoder to locate this GOB. Furthermore, the GOB header contains information which allows the decoding process to be restarted (i.e., resynchronize the decoder to the bitstream and reset all predictively coded data).
The GOB approach to resynchronization is based on spatial resynchronization. That is, once a particular macroblock location is reached in the encoding process, a resynchronization marker is inserted into the bitstream. A potential problem with this approach is that since the encoding process is variable rate, these resynchronization markers will most likely be unevenly spaced throughout the bitstream. Therefore, certain portions of the scene, such as high motion areas, will be more susceptible to errors, which will also be more difficult to conceal.
The video packet approach adopted by MPEG-4 is based on providing periodic resynchronization markers throughout the bitstream. In other words, the length of the video packets are not based on the number of macroblocks, but instead on the number of bits contained in that packet. If the number of bits contained in the current video packet exceeds a predetermined threshold, then a new video packet is created at the start of the next macroblock.
A resynchronization marker is used to distinguish the start of a new video packet. This marker is distinguishable from all possible VLC codewords as well as the VOP start code. Header information is also provided at the start of a video packet. Contained in this header is the information necessary to restart the decoding process and includes: the macroblock number of the first macroblock contained in this packet and the quantization parameter necessary to decode that first macroblock. The macroblock number provides the necessary spatial resynchronization while the quantization parameter allows the differential decoding process to be resynchronized.
Also included in the video packet header is the header extension code. The HEC is a single bit that, when enabled, indicates the presence of additional resynchronization information; including modular time base, VOP temporal increment, VOP prediction type, and VOP F code. This additional information is made available in case the VOP header has been corrupted.
It should be noted that when utilizing the error resilience tools within MPEG-4, some of the compression efficiency tools are modified. For example, all predictively encoded information must be confined within a video packet so as to prevent the propagation of errors.
In conjunction with the video packet approach to resynchronization, a second method called fixed interval synchronization has also been adopted by MPEG-4. This method requires that VOP start codes and resynchronization markers (i.e., the start of a video packet) appear only at legal fixed interval locations in the bitstream. This helps avoiding the problems associated with start codes emulations. That is, when errors are present in a bitstream it is possible for these errors to emulate a VOP start code. In this case, when fixed interval synchronization is utilized the decoder is only required to search for a VOP start code at the beginning of each fixed interval. The fixed interval synchronization method extends this approach to be any predetermined interval.
After synchronization has been reestablished, data recovery tools attempt to recover data that in general would be lost. These tools are not simply error correcting codes, but instead techniques that encode the data in an error resilient manner. For instance, one particular tool that has been endorsed by the Video Group is Reversible Variable Length Codes (RVLC). In this approach, the variable length codewords are designed such that they can be read both in the forward as well as the reverse direction.
An example illustrating the use of a RVLC is given in Figure 14. Generally, in a situation such as this, where a burst of errors has corrupted a portion of the data, all data between the two synchronization points would be lost. However, as shown in the Figure, an RVLC enables some of that data to be recovered. It should be noted that the parameters, QP and HEC shown in the Figure, represent the fields reserved in the video packet header for the quantization parameter and the header extension code, respectively.
Figure 14 - example of Reversible Variable Length Code
Error concealment is an extremely important component of any error robust video codec. Similar to the error resilience tools discussed above, the effectiveness of an error concealment strategy is highly dependent on the performance of the resynchronization scheme. Basically, if the resynchronization method can effectively localize the error then the error concealment problem becomes much more tractable. For low bitrate, low delay applications the current resynchronization scheme provides very acceptable results with a simple concealment strategy, such as copying blocks from the previous frame.
In recognizing the need to provide enhanced concealment capabilities, the Video Group has developed an additional error resilient mode that further improves the ability of the decoder to localize an error.
Specifically, this approach utilizes data partitioning by separating the motion and the texture. This approach requires that a second resynchronization marker be inserted between motion and texture information. If the texture information is lost, this approach utilizes the motion information to conceal these errors. That is, due to the errors the texture information is discarded, while the motion is used to motion compensate the previous decoded VOP.
A major application area is likely to be interactive web-based video. Software that provides live MPEG-4 video on a web-page has already been demonstrated. There is much room for applications to make use of MPEG-4's object-based characteristics. The binary and grayscale shape-coding tools allow arbitrary-shaped video objects to be composed together with text and graphics. Doing so, a rich interactive experience for web-based presentations and advertising can be provided; this same scenario also applies to set-top box applications. Additionally, it is possible to make use of scalability tools to allow for a smooth control of user experience with terminal and data link capabilities.
MPEG-4 video has already been used to encode video captures with a hand-held camera. This form of application is likely to grow in popularity with its fast and easy transfer to a web-page, and may also make use of MPEG-4 still-texture mode for still frame capture. The games market is another area where the application of MPEG-4 video, still-texture, interactivity and SNHC shows much promise, with 3-D texture mapping of still images, live video, or extended pre-recorded video sequences enhancing the player experience. Adding live video of users adds to the user experience multi-player 3-D games, as does use of arbitrary-shaped video, where transparency could be combined artistically with 3-D video texture mapping.
The flexibility of MPEG-4 video coding encourages many more applications like this one.
In addition to providing support for coding individual objects, MPEG-4 also provides facilities to compose a set of such objects into a scene. The necessary composition information forms the scene description, which is coded and transmitted together with the media objects. Starting from VRML (the Virtual reality Modeling Language), MPEG has developed a binary language for scene description called BIFS. BIFS stands for BInary Format for Scenes.
In order to facilitate the development of authoring, manipulation and interaction tools, scene descriptions are coded independently from streams related to primitive media objects. Special care is devoted to the identification of the parameters belonging to the scene description. This is done by differentiating parameters that are used to improve the coding efficiency of an object (e.g., motion vectors in video coding algorithms), and the ones that are used as modifiers of an object (e.g., the position of the object in the scene). Since MPEG-4 should allow the modification of this latter set of parameters without having to decode the primitive media objects themselves, these parameters are placed in the scene description and not in primitive media objects.
The following list gives some examples of the information described in a scene description.
How objects are grouped together: An MPEG-4 scene follows a hierarchical structure, which can be represented as a directed acyclic graph. Each node of the graph is a media object, as illustrated in Figure 15 (note that this tree refers back to Figure 1). The tree structure is not necessarily static; node attributes (e.g., positioning parameters) can be changed while nodes can be added, replaced, or removed.
Figure 15- Logical structure of a scene
How objects are positioned in space and time: In the MPEG-4 model, audiovisual objects have both a spatial and a temporal extent. Each media object has a local coordinate system. A local coordinate system for an object is one in which the object has a fixed spatio-temporal location and scale. The local coordinate system serves as a handle for manipulating the media object in space and time. Media objects are positioned in a scene by specifying a coordinate transformation from the object’s local coordinate system into a global coordinate system defined by one more parent scene description nodes in the tree.
Attribute Value Selection: Individual media objects and scene description nodes expose a set of parameters to the composition layer through which part of their behavior can be controlled. Examples include the pitch of a sound, the color for a synthetic object, activation or deactivation of enhancement information for scaleable coding, etc.
Other transforms on media objects: As mentioned above, the scene description structure and node semantics are heavily influenced by VRML, including its event model. This provides MPEG-4 with a very rich set of scene construction operators, including graphics primitives that can be used to construct sophisticated scenes.
MPEG-4 allows for user interaction with the presented content. This interaction can be separated into two major categories: client-side interaction and server-side interaction. Client-side interaction involves content manipulation, which is handled locally at the end-user’s terminal, and can take several forms. In particular, the modification of an attribute of a scene description node, e.g., changing the position of an object, making it visible or invisible, changing the font size of a synthetic text node, etc., can be implemented by translating user events. A user event can be a mouse clicks or keyboard command) to scene description updates. The MPEG-4 terminal can process the commands in exactly the same way as if they originated from the original content source. As a result, this type of interaction does not require standardization.
Other forms of client-side interaction require support from the scene description syntax, and are specified by the standard. The use of the VRML event structure provides a rich model on which content developers can create compelling interactive content.
Server-side interaction involves content manipulation that occurs at the transmitting end, initiated by a user action. This, of course, requires that a back-channel is available.
MPEG-4 provides mechanisms for protection of intellectual property rights (IPR), as outlined in section 1.5. This is achieved by supplementing the coded media objects with an optional Intellectual Property Identification (IPI) data set, carrying information about the contents, type of content and (pointers to) rights holders. The data set, if present, is part of an elementary stream descriptor that describes the streaming data associated to a media object. The number of data sets to be associated with each media object is flexible; different media objects can share the same data sets or have separate data sets. The provision of the data sets allows the implementation of mechanisms for audit trail, monitoring, billing, and copy protection.
Next to identifying rights, each of the wide range of MPEG-4 applications has a set of requirements regarding protection of the information it manages. These applications can have different security requirements. For some applications, users exchange information that has no intrinsic value but that must still be protected to preserve various rights of privacy. For other applications, the managed information has great value to its creator and/or distributors requiring high-grade management and protection mechanisms. The implication is that the design of the IPMP framework must consider the complexity of the MPEG-4 standard and the diversity of its applications. This IPMP framework leaves the details of IPMP systems designs in the hands of applications developers. The level and type of management and protection required depends on the content’s value, complexity, and the sophistication of the associated business models.
The approach taken allows the design and use of domain-specific IPMP systems (IPMP-S). While MPEG-4 does not standardize IPMP systems themselves, it does standardize the MPEG-4 IPMP interface. This interface consists of IPMP-Descriptors (IPMP-Ds) and IPMP-Elementary Streams (IPMP-ES).
IPMP-Ds and IPMP-ESs provide a communication mechanism between IPMP systems and the MPEG-4 terminal. Certain applications may require multiple IPMP systems. When MPEG-4 objects require management and protection, they have IPMP-Ds associated with them. These IPMP-Ds indicate which IPMP systems are to be used and provide information to these systems about how to manage and protect the content. (See Figure 16)
Besides enabling owners of intellectual property to manage and protect their assets, MPEG-4 provides a mechanism to identify those assets via the Intellectual Property Identification Data Set (IPI Data Set). This information can be used by IPMP systems as input to the management and protection process.
Figure 16 - The IPMP Interfaces within an MPEG-4 System
MPEG-4 allows attaching information to objects about their content. Users of the standard can use this ‘OCI’ (Object Content Information) data stream to send textual information along with MPEG-4 content. It is also possible to classify content according to pre-defined tables, which will be defined outside of MPEG. Further possibilities are giving unique labels to content, and storing camera parameters.
This section contains, in an itemized fashion, the major functionalities that the different parts of the MPEG-4 Standard offers in the finalized MPEG-4 Version 1. Description of the functionalities can be found above.
DMIF supports the following functionalities:
MPEG-4 Audio facilitates a wide variety of applications which could range from intelligible speech to high quality multichannel audio, and from natural sounds to synthesized sounds. In particular, it supports the highly efficient representation of audio objects consisting of:
Examples of additional functionality are speed control and pitch change for speech signals and scalability in terms of bitrate, bandwidth, error robustness, complexity, etc. as defined below.
The MPEG-4 Visual standard will allow the hybrid coding of natural (pixel based) images and video together with synthetic (computer generated) scenes. This will, for example, allow the virtual presence of videoconferencing participants. To this end, the Visual standard will comprise tools and algorithms supporting the coding of natural (pixel based) still images and video sequences as well as tools to support the compression of synthetic 2-D and 3-D graphic geometry parameters (i.e. compression of wire grid parameters, synthetic text).
The subsections below give an itemized overview of functionalities that the tools and algorithms of the MPEG-4 visual standard will support.
The following formats and bitrates will be supported by MPEG-4 Version 1:
An alpha plane defines the ‘transparency’ of an object, which is not necessarily uniform. Multilevel alpha maps are frequently used to blend different layers of image sequences. Other applications that benefit from associated binary alpha maps with images are content based image representations for image databases, interactive games, surveillance, and animation. Efficient techniques are provided, that allow efficient coding of binary as well as gray scale alpha planes. A binary alpha map defines whether or not a pixel belongs to an object. It can be ‘on’ or ‘off’. A gray scale map offers the possibility to define the exact transparency of each pixel.
Error resilience will be supported to assist the access of image and video over a wide range of storage and transmission media. This includes the useful operation of image and video compression algorithms in error-prone environments at low bit-rates (i.e., less than 64 Kbps). Tools are provided which address both the band limited nature and error resiliency aspects for access over wireless networks.
The ‘Face Animation’ part of the standard allow sending parameters that calibrate and animate synthetic faces. These models themselves are not standardized by MPEG-4, only the parameters are.
In October 1998, the first set of MPEG-4 Standards was frozen. Work on MPEG-4 continued for Version 2, which will add tools to the MPEG-4 Standard. Version 2 was frozen in December 1999. Existing tools and profiles from Version 1 will not be replaced in Version 2; technology will be added to MPEG-4 in the form of new profiles. Figure 17 below depicts the relationship between the two versions. Version 2 is a backward compatible extension of Version 1.
Figure 17 - relation between MPEG-4 Versions
Version 2 builds on Version 1 of MPEG-4. The Systems layer of Version 2 is backward compatible with Version 1. In the area of Audio and Visual, Version 2 will add Profiles to Version 1. The work on MPEG-4 does not stop after MPEG-4; more functionality will be added, albeit in particular, well-defined areas. The same principle applies, and new tools will find their way in the standard in the form of new Profiles. This means that existing systems will always remain compliant, because Profiles will not be changed in retrospect.
Version 2 of the MPEG-4 systems extends Version 1 to cover issues like extended BIFS functionalities, and Java (MPEG-J) support. Version 2 also specifies a file format to store MPEG-4 content. Below the new elements will be briefly introduced.
The version 2 BIFS (Advanced BIFS) includes the following new functionalities:
The MP4 file format is designed to contain the media information of an MPEG-4 presentation in a flexible, extensible format which facilitates interchange, management, editing, and presentation of the media. This presentation may be ‘local’ to the system containing the presentation, or may be via a network or other stream delivery mechanism (a TransMux). The file format is designed to be independent of any particular delivery protocol while enabling efficient support for delivery in general. The design is based on the QuickTime® format from Apple Computer Inc.
The following diagram gives an example of a simple interchange file, containing three streams.
The MP4 file format is composed of object-oriented structures called ‘atoms’. A unique tag and a length identify each atom. Most atoms describe a hierarchy of metadata giving information such as index points, durations, and pointers to the media data. This collection of atoms is contained in an atom called the ‘movie atom’. The media data itself is located elsewhere; it can be in the MP4 file, contained in one or more ‘mdat’ or media data atoms, or located outside the MP4 file and referenced via URL’s.
The following diagram shows a more complex file with external media data:
The following diagram shows the container relationship between the different objects or atoms:
The file format is a streamable format, as opposed to a streaming format. That is, the file format does not define an on-the-wire protocol, and is never actually streamed over a transmission medium. Instead, metadata in the file known as ‘hint tracks’ provide instructions, telling a server application how to deliver the media data over a particular delivery protocol. There can be multiple hint tracks for one presentation, describing how to deliver over various delivery protocols. In this way, the file format facilitates streaming without ever being streamed directly
The following diagram shows the container relationship with RTP protocol hint tracks to stream a simple video movie:
The metadata in the file, combined with the flexible storage of media data, allows the MP4 format to support streaming, editing, local playback, and interchange of content, thereby satisfying the requirements for the MPEG4 Intermedia format.
The MPEG-J is a programmatic system (as opposed to the parametric system offered by MPEG-4 Version 1) which specifies API for interoperation of MPEG-4 media players with Java code. By combining MPEG-4 media and safe executable code, content creators may embed complex control and data processing mechanisms with their media data to intelligently manage the operation of the audio-visual session. A block diagram of the MPEG-J player in an MPEG-4 system player environment is shown in Figure 18. The lower half of this drawing depicts the parametric MPEG-4 System player also referred to as the Presentation Engine. The MPEG-J subsystem controlling the Presentation Engine, also referred to as the Application Engine, is depicted in the upper half of Figure 18.
The Java application is delivered as a separate elementary stream to the MPEG-4 terminal. There it will be directed to the MPEG-J run time environment, from where the MPEG-J program will have access to the various components and data of the MPEG-4 player, in addition to .the basic packages of the language (java.lang, java.io, java.util). MPEG-J specifically does not support downloadable decoders.
For the above-mentioned reason, the group has defined a set of APIs with different scopes. For Scene graph API the objective is to provide access to the scene graph: to inspect the graph, to alter nodes and their fields, and to add and remove nodes within the graph. The Resource Manager API is used for regulation of performance: it provides a centralized facility for managing resources. The Terminal Capability API is used when program execution is contingent upon the terminal configuration and its capabilities, both static (that do not change during execution) and dynamic. Media Decoders API allow the control of the decoders that are present in the terminal. The Network API provides a way to interact with the network, being compliant to the MPEG-4 DMIF Application Interface. Complex applications and enhanced interactivity are possible with these basic packages.
Figure 18 - Location of interfaces in the architecture of an MPEG-J enabled MPEG-4 System
MPEG-4 Visual Version 2 adds technology in the following areas:
MPEG-4 V.2 improves the motion estimation and compensation of rectangular and arbitrary shaped objects and the texture coding of arbitrary shaped objects significantly. In the area of motion estimation and compensation two new techniques are introduced:
In the area of texture coding the shape-adaptive DCT (SA-DCT) improves the coding efficiency of arbitrary shaped objects. The SA-DCT algorithm is based on predefined orthonormal sets of one-dimensional DCT basis functions.
Subjective evaluation tests within MPEG have shown that the combination of these techniques can result in a bitstream saving of up to 50% compared with the version 1, depending on content type and datarate.
There are several scalable coding schemes in MPEG-4 Visual: spatial scalability, temporal scalability and object-based spatial scalability. Spatial scalability supports changing the texture quality (SNR and spatial resolution). Object-based spatial scalability extends the 'conventional' types of scalability towards arbitrary shape objects, so that it can be used in conjunction with other object-based capabilities. Thus, a very flexible content-based scaling of video information can be achieved. This makes it possible to enhance SNR, spatial resolution, shape accuracy, etc, only for objects of interest or for a particular region, which can een be done dynamically at play-time.
A new technique, called NEWPRED (for 'new prediction') provides a fast error recovery in real-time coding applications. It uses an upstream channel from the decoder to the encoder. The encoder switches the reference frames adaptively according to the error conditions of the network. NEWPRED does not use intra refresh and it provides the high coding efficiency. In addition, dynamic resolution conversion (DRC) is introduced, a way to stabilize the transmission buffering delay by minimizing the jitter of the amount of the coded output bits per VOP. Large frame skips are also prevented and the encoder can control the temporal resolution even in highly active scenes.
These techniques have been proven to work under stressful error conditions:
The following three new tools for coding of textures and still images are provided in V.2:
In MPEG-4 Video Version 1 up to one alpha channel per Video Object Layer is supported and in total three shape types are defined. All the three shape types, i.e. binary shape, constant shape and gray scale shape, indicate the transparency of the video object. With this definition MPEG-4 cannot support things like Multiview Video Objects in an efficient way. Therefore in Version 2 the use of multiple alpha channels to transmit auxiliary components is introduced.
The basic idea is that the gray scale shape is not only used to describe the transparency of the video object, but can be defined in a more general way. A gray scale shape may e.g. represent:
All the alpha channels can be encoded by the shape coding tools, i.e. the binary shape coding tool and the gray scale shape coding tool which employs a motion-compensated DCT, and usually have the same shape and resolution as the texture of the video object.
As an example for employing multiple auxiliary components, the usage of the disparity shape for Multiview Video Objects is described below.
The general concept is to limit the number of pixels that have to be encoded by analysis of the correspondences between the particular views of an object available at the encoder side. All areas of an object that are visible within more than one camera view are encoded only once with the highest possible resolution. The disparity correspondences can be estimated from the original views to reconstruct all the areas that were excluded from encoding by use of disparity-compensated projection. One or two auxiliary components can be allocated to encode the disparity maps, indicating the correspondences between the pixels within multiple views.
We denominate the areas which are retained for encoding from each of the particular camera views as the "areas of interest" (AOI). These AOIs can now simply be defined as MPEG-4 video objects, and are encoded with their associated disparity values. Due to possible reflection effects in the different views, but as well due to color or exposure deviations between the cameras, borders between the areas which are to be reconstructed from different original views might become visible. To circumvent this problem, it is necessary to preprocess pixels near the borders of an AOI, such that a smooth transition is achieved by interpolating pixels from different adjacent views within a transition area.
To reconstruct different viewpoints from the texture surface disparity-compensated projection is performed from the texture data within the particular AOIs, with disparities from the disparity map available within the auxiliary component decoded from the MPEG-4 video stream. Each of the AOIs is processed separately and then the projected images from all AOIs are assembled to obtain the final view towards the video object from the selected viewpoint. This procedure can be employed for a system with two cameras with parallel setup, but is also extendable to convergent and multiple-camera cases as well.
The Body is an object capable of producing virtual body models and animations in form of a set of 3-D polygonal meshes ready for rendering. Two sets of parameters are defined for the body: Body Definition Parameter (BDP) set, and Body Animation Parameter (BAP) set. The BDP set defines the set of parameters to transform the default body to a customized body with its body surface, body dimensions, and (optionally) texture. The Body Animation Parameters (BAPs), if correctly interpreted, will produce reasonably similar high level results in terms of body posture and animation on different body models, without the need to initialize or calibrate the model.
Upon construction, the Body object contains a generic virtual human body with the default posture. This body can already be rendered. It is also immediately capable of receiving the BAPs from the bitstream, which will produce animation of the body. If BDPs are received, they are used to transform the generic body into a particular body determined by the parameters contents. Any component can be null. A null component is replaced by the corresponding default component when the body is rendered. The default posture is defined by standing posture. This posture is defined as follows: the feet should point to the front direction, the two arms should be placed on the side of the body with the palm of the hands facing inward. This posture also implies that all BAPs have default values.
No assumption is made and no limitation is imposed on the range of motion of joints. In other words the human body model should be capable of supporting various applications, from realistic simulation of human motions to network games using simple human-like models. The work on Body Animation includes the assessment of the emerging standard as applied to hand signing for the listening-impaired.
The Body Animation standard is being developed by MPEG in concert with the Humanoid Animation Working Group within the VRML Consortium, with the objective of achieving consistent conventions and control of body models which are being established by H-Anim.
Version 2 MPEG-4 provides a suite of tools for coding 3-D polygonal meshes. Polygonal meshes are widely used as a generic representation of 3-D objects. The underlying technologies compress the connectivity, geometry, and properties such as shading normals, colors and texture coordinates of 3-D polygonal meshes. Capabilities for 3-D mesh coding include:
MPEG-4 Audio Version 2 is an extension to MPEG-4 Audio Version 1. It adds new tools and functionalities to the MPEG-4 Standard, while none of the existing tools of Version 1 is replaced. The following additional functionalities are provided by MPEG-4 Audio Version 2:
The error robustness tools provide improved performance on error-prone transmission channels. They can be distinguished into codec specific error resilience tools and a common error protection tool.
Improved error robustness for AAC is provided by a set of error resilience tools. These tools reduce the perceived deterioration of the decoded audio signal that is caused by corrupted bits in the bit stream. The following tools are provided to improve the error robustness for several parts of an AAC frame:
Improved error robustness capabilities for all coding tools are provided through the error resilient bit stream payload syntax. It allows advanced channel coding techniques, which can be adapted to the special needs of the different coding tools. This error resilient bit stream payload syntax is mandatory for all Version 2 object types.
The error protection tool (EP tool) provides error protection for all MPEG-4 Audio version 2 audio objects with flexible configuration applicable for wide range of channel error conditions. The main features of the EP tool are as follows:
MPEG-4 Audio version 2 coding algorithms provide a classification of each bit stream field according to its error sensitivity. Based on this, the bit stream is divided into several classes, which can be separately protected by the EP tool, such that more error sensitive parts are protected more strongly.
While the MPEG-4 General Audio Coder provides very efficient coding of general audio signals at low bit rates, it has an algorithmic encoding/decoding delay of up to several 100ms and is thus not well suited for applications requiring low coding delay, such as real-time bi-directional communication. As an example, for the General Audio Coder operating at 24 kHz sampling rate and 24 kbit/s, this results in an algorithmic coding delay of about 110 ms plus up to additional 210 ms for the use of the bit reservoir. To enable coding of general audio signals with an algorithmic delay not exceeding 20 ms, MPEG-4 Version 2 specifies a Low-Delay Audio Coder which is derived from MPEG-2/4 Advanced Audio Coding (AAC). . Compared to speech coding schemes, this coder allows compression of general audio signal types, including music, at a low delay. It operates at up to 48 kHz sampling rate and uses a frame length of 512 or 480 samples, compared to 1024 or 960 samples used in standard MPEG-2/4 AAC. Also the size of the window used in the analysis and synthesis filter bank is reduced by a factor of 2. No block switching is used to avoid the "look-ahead'' delay due to the block switching decision. To reduce pre-echo artifacts in case of transient signals, window shape switching is provided instead. For non-transient parts of the signal a sine window is used, while a so-called low overlap window is used in case of transient signals. Use of the bit reservoir is minimized in the encoder in order to reach the desired target delay. As one extreme case, no bit reservoir is used at all. Verification tests have shown that the reduction in coding delay comes at a very moderate cost in compression performance.
Bit rate scalability, also known as embedded coding, is a very desirable functionality. The General Audio Coder of Version 1 supports large step scalability where a base layer bit stream can be combined with one or more enhancement layer bit streams to utilize a higher bit rate and thus obtain a better audio quality. In a typical configuration, a 24 kbit/s base layer and two 16 kbit/s enhancement layers could be used, permitting decoding at a total bit rate of 24 kbit/s (mono), 40 kbit/s (stereo), and 56 kbit/s (stereo). Due to the side information carried in each layer, small bit rate enhancement layers are not efficiently supported in Version 1. To address this problem and to provide efficient small step scalability for the General Audio Coder, the Bit-Sliced Arithmetic Coding (BSAC) tool is available in Version 2. This tool is used in combination with the AAC coding tools and replaces the noiseless coding of the quantized spectral data and the scale factors. BSAC provides scalability in steps of 1 kbit/s per audio channel, i.e. 2 kbit/s steps for a stereo signal. One base layer bit stream and many small enhancement layer bit streams are used. The base layer contains the general side information, specific side information for the first layer and the audio data of the first layer. The enhancement streams contain only the specific side information and audio data for the corresponding layer. To obtain fine step scalability, a bit-slicing scheme is applied to the quantized spectral data. First the quantized spectral values are grouped into frequency bands. Each of these groups contains the quantized spectral values in their binary representation. Then the bits of a group are processed in slices according to their significance. Thus first all most significant bits (MSB) of the quantized values in a group are processed, etc. These bit-slices are then encoded using an arithmetic coding scheme to obtain entropy coding with minimal redundancy. Various arithmetic coding models are provided to cover the different statistics of the bit-slices. The scheme used to assign the bit-slices of the different frequency bands to the enhancement layer is constructed in a special way. This ensures that, with an increasing number of enhancement layers utilized by the decoder, providing more of the less significant bits refines quantized spectral data. But also providing bit-slices of the spectral data in higher frequency bands increases the bandwidth.
Verification tests have shown that the scalability aspect of this tool performs well over a wide range of rates. At the highest rate it is as good as AAC main profile operating at the same rate, while at the lowest rate the scalability function requires a moderate overhead relative to AAC main profile operating at the same rate.
The Parametric Audio Coding tools combine very low bit rate coding of general audio signals with the possibility of modifying the playback speed or pitch during decoding without the need for an effects processing unit. In combination with the speech and audio coding tools of Version 1, improved overall coding efficiency is expected for applications of object based coding allowing selection and/or switching between different coding techniques.
Parametric Audio Coding uses the Harmonic and Individual Lines plus Noise (HILN) technique to code general audio signals at bit rates of 4 kbit/s and above using a parametric representation of the audio signal. The basic idea of this technique is to decompose the input signal into audio objects, which are described by appropriate source models and represented by model parameters. Object models for sinusoids, harmonic tones, and noise are utilized in the HILN coder.
This approach allows to introduce a more advanced source model than just assuming a stationary signal for the duration of a frame, which motivates the spectral decomposition used e.g. in the MPEG-4 General Audio Coder. As known from speech coding, where specialized source models based on the speech generation process in the human vocal tract are applied, advanced source models can be advantageous in particular for very low bit rate coding schemes.
Due to the very low target bit rates, only the parameters for a small number of objects can be transmitted. Therefore a perception model is employed to select those objects that are most important for the perceptual quality of the signal.
In HILN, the frequency and amplitude parameters are quantized according to the "just noticeable differences" known from psychoacoustics. The spectral envelope of the noise and the harmonic tone is described using LPC modeling as known from speech coding. Correlation between the parameters of one frame and between consecutive frames is exploited by parameter prediction. The quantized parameters are finally entropy coded and multiplexed to form a bit stream.
A very interesting property of this parametric coding scheme arises from the fact that the signal is described in terms of frequency and amplitude parameters. This signal representation permits speed and pitch change functionality by simple parameter modification in the decoder. The HILN parametric audio coder can be combined with MPEG-4 parametric speech coder (HVXC) to form an integrated parametric coder covering a wider range of signals and bit rates. This integrated coder supports speed and pitch change. Using a speech/music classification tool in the encoder, it is possible to automatically select the HVXC for speech signals and the HILN for music signals. Such automatic HVXC/HILN switching was successfully demonstrated and the classification tool is described in the informative Annex of the Version 2 standard.
Verification test have shown that HILN coding has performance comparable to other MPEG-4 coding technology operating at similar bit rates while providing the additional capability of independent audio signal speed or pitch change when decoding. The test has also shown that the scalable HILN coder provides quality comparable to that of a fixed-rate HILN coder at the same bit rate.
The silence compression tool reduces the average bit rate thanks to a lower bit rate compression for silence. In the encoder, a voice activity detector is used to distinguish between regions with normal speech activity and those with silence or background noise. During normal speech activity, the CELP coding as in Version 1 is used. Otherwise a Silence Insertion Descriptor (SID) is transmitted at a lower bit rate. This SID enables a Comfort Noise Generator (CNG) in the decoder. The amplitude and spectral shape of this comfort noise is specified by energy and LPC parameters similar as in a normal CELP frame. These parameters are an optional part of the SID and thus can be updated as required.
The Error Resilient (ER) HVXC object is supported by the Parametric speech coding (ER HVXC) tools, which provides fixed bit-rate modes(2.0-4.0kbps) and variable bit-rate mode(<2.0kbps, <4.0kbps) both in a scalable and non-scalable scheme. In the Version 1 HVXC, variable bit rate mode of 2.0 kbit/s maximum is already supported and the variable bit rate mode of 4.0 kbit/s maximum is additionally supported in Version 2 ER HVXC. In the variable bit rate modes, non-speech parts are detected in unvoiced signals, and a smaller number of bits is used for these non-speech parts to reduce the average bit rate. ER HVXC provides communications-quality to near-toll-quality speech in the 100-3800 Hz band at 8kHz sampling rate. When the variable bit-rate mode is allowed, operation at lower average bit-rate is possible. Coded speech with variable bit-rate mode at typical bit-rate of 1.5kbps average, and at typical bit-rate of 3.0kbps average has essentially the same quality as 2.0 kbps fixed rate and 4.0 kbps fixed rate respectively. The functionality of pitch and speed change during decoding is supported for all modes. ER HVXC has the syntax with the error sensitivity classes to be used with the EP-Tool, and the error concealment functionality are supported for the use for error-prone channel like mobile communication channels. The ER HVXC speech coder targets applications from mobile and satellite communications, to Internet telephony, to packaged media and speech databases.
The Environmental Spatialization tools enable composition of an audio scene with more natural sound source and sound environment modeling than is possible in Version 1. Both, a physical and a perceptual approach to spatialization are supported. _ The physical approach is based on a description of the acoustical properties of the environment (e.g. room geometry, material properties, position of sound source) and can be used in applications like 3-D virtual reality. _The perceptual approach on the other hand permits a high level perceptual description of the audio scene based on parameters similar to those of a reverberation effects unit. Thus, the audio and the visual scene can be composed independently as usually required by applications like movies. Although the Environmental Spatialization tools are related to audio, they are part of the BInary Format for Scene description (BIFS) in MPEG-4 Systems and are referred to as Advanced AudioBIFS.
The back channel allows a request of client and/or client terminal to server. With this capability, interactivity can be achieved. In MPEG-4 System, the need for an up-stream channel (back channel) is signaled to the client terminal by supplying an appropriate elementary stream descriptor declaring the parameters for that stream. The client terminal opens this upstream channel in a similar manner as it opens the down-stream channels. The entities (e.g. media encoders & decoders) that are connected through an upstream channel are known from the parameters in its elementary stream descriptor and from the association of the elementary stream descriptor to a specific object descriptor. In MPEG-4 Audio, the back channel allows feedback for bit rate adjustment, the scalability and error protection adaptation.
The MPEG-4 Audio transport stream defines a mechanism to transport MPEG-4 Audio streams without using MPEG-4 Systems and is dedicated for audio-only applications. The transport mechanism uses a two-layer approach, namely a multiplex layer and a synchronization layer. The multiplex layer (Low-overhead MPEG-4 Audio Transport Multiplex: LATM) manages multiplexing of several MPEG-4 Audio payloads and audio specific configuration information. The synchronization layer specifies a self-synchronized syntax of the MPEG-4 Audio transport stream which is called Low Overhead Audio Stream (LOAS). The Interface format to a transmission layer depends on the conditions of the underlying transmission layer.
The Delivery efforts for MPEG-4 Version 2 will consider the following issues.
DMIF SRM allows the collection of accounting logs so that the revenues collected from the peers in a DMIF session are properly disbursed to those providers that supply the resources within that session.
MPEG-4 provides a large and rich set of tools for the coding of audio-visual objects. In order to allow effective implementations of the standard, subsets of the MPEG-4 Systems, Visual, and Audio tool sets have been identified, that can be used for specific applications. These subsets, called ‘Profiles’, limit the tool set a decoder has to implement. For each of these Profiles, one or more Levels have been set, restricting the computational complexity. The approach is similar to MPEG-2, where the most well known Profile/Level combination is ‘Main Profile @ Main Level’. A Profile@Level combination allows:
Profiles exist for various types of media content (audio, visual, and graphics) and for scene descriptions. MPEG does not prescribe or advise combinations of these Profiles, but care has been taken that good matches exist between the different areas.
The visual part of the standard provides profiles for the coding of natural, synthetic, and synthetic/natural hybrid visual content. There are five profiles for natural video content:
The profiles for synthetic and synthetic/natural hybrid visual content are:
Version 2 adds the following Profiles for natural video:
The Version 2 profiles for synthetic and synthetic/natural hybrid visual content are:
Four Audio Profiles have been defined in MPEG-4 V.1:
Another four Profiles were added in MPEG-4 V.2:
Graphics Profiles define which graphical and textual elements can be used in a scene. These profiles are defined in the Systems part of the standard:
Scene Description Profiles, defined in the Systems part of the standard, allow audiovisual scenes with audio-only, 2-dimensional, 3-dimensional or mixed 2-D/3-D content. The 3-D Profile is called ‘VRML’, as it optimizes interworking with VRML material:
Two MPEG-J Profiles exist:
The Object Descriptor Profile includes the following tools:
Currently, only one profile is defined that includes all these tools. The main reason for defining this profiles is not subsetting the tools, but rather defining levels for them. This applies especially to the Sync Layer tool, as MPEG-4 allows multiple time bases to exist. In the context of Levels for this Profile, restrictions can be defined, e.g. to allow only a single time base.
MPEG carries out verification tests to check whether the standard delivers what it promises.
The first formal tests on MPEG-4 Audio codecs were completed, based on collaboration between MPEG and the NADIB (Narrowband Digital Audio Broadcasting) Group. These tests explored the performance of speech and music coders working in the bitrate range 6 kb/s to 24 kb/s, including some scaleable codec options. The results show that a significant improvement in quality can be offered in relation to conventional analogue AM broadcasting and that scaleable coders offer superior performance to simulcast operations.
More verification tests are being prepared. For Video, error robustness, content-based coding and scalability will be evaluated. For Audio: tests are prepared for speech coding and audio on the Internet.
The Moving Picture Coding Experts Group (MPEG) is a working group of ISO/IEC in charge of the development of international standards for compression, decompression, processing, and coded representation of moving pictures, audio and their combination.
The purpose of MPEG is to produce standards. The first two standards produced by MPEG were:
MPEG is has recently finalized MPEG-4 Version 1, a standard for multimedia applications, that will officially reach the status of International Standard in February 1999, with the ISO number 14496.
MPEG has also started work on a new standard known as MPEG-7: a content representation standard for information search, scheduled for completion in Fall 2001. The Call for Proposals was issued in October 1998.
MPEG-1 has been a very successful standard. It is the de-facto form of storing moving pictures and audio on the World Wide Web and is used in millions of Video CDs. Digital Audio Broadcasting (DAB) is a new consumer market that makes use of MPEG-1 audio coding.
MPEG-2 has been the timely response for the satellite broadcasting and cable television industries in their transition from analogue to digital. Millions of set-top boxes incorporating MPEG-2 decoders have been sold in the last 3 years.
Since July 1993 MPEG is working on its third standard, called MPEG-4. MPEG considers of vital importance to define and maintain, without slippage, a work plan. This is the MPEG-4 Version 1 work plan:
| Part | Title | WD | CD | FCD | FDIS | IS |
| 1 | Systems | 97/11 | 98/03 | 98/10 | 99/04 | |
| 2 | Visual | 97/11 | 98/03 | 98/10 | 99/04 | |
| 3 | Audio | 97/11 | 98/03 | 98/10 | 99/04 | |
| 4 | Conformance Testing | 97/10 | 98/12 | 99/07 | 99/12 | 00/02 |
| 5 | Reference Software | 97/11 | 98/03 | 99/03 | 99/05 | |
| 6 | Delivery Multimedia Integration Framework (DMIF) | 97/07 | 97/11 | 98/03 | 98/10 | 99/04 |
| 1/Amd 1 | Systems Extensions (v.2) | 97/10 | 99/03 | 99/07 | 99/12 | 00/02 |
| 2/Amd 1 | Visual Extensions (v.2) | 97/10 | 99/03 | 99/07 | 99/12 | 00/02 |
| 3/Amd 1 | Audio Extensions (v.2) | 97/10 | 99/03 | 99/07 | 99/12 | 00/02 |
| 4/Amd1 | Conformance Testing Extensions (v.2) | 98/12 | 99/12 | 00/07 | 00/12 | 01/02 |
| 5/Amd 1 | Reference Software Extensions (v.2) | 97/10 | 99/07 | 99/12 | 00/03 | 00/05 |
| 6/Amd 1 | DMIF Extensions (v.2) | 97/10 | 99/03 | 99/07 | 99/12 | 00/02 |
Table 1 - MPEG-4 work plan (NB: The abbreviations are explained below)
Because of the complexity of the work item, it took 2 years before a satisfactory definition of the scope could be achieved and half a year more before a first call for proposals could be issued. This call, like all MPEG calls, was open to all interested parties, no matter whether they were within or outside of MPEG. It requested technology that proponents felt could be considered by MPEG for the purpose of the developing the MPEG-4 standard. After that first call, other calls were issued for other technology areas.
The proposals of technology received were assessed and, if found promising, incorporated in the so-called Verification Models (VMs). A VM describes, in text and some sort of programming language, the operation of encoder and decoder. VMs are used to carry out simulations with the aim to optimize the performance of the coding schemes.
Because (next to the envisaged hardware environments for MPEG-4) software platforms are gaining in importance for multimedia standards, MPEG decided to maintain software implementations of the different standard parts. These can be used for the purpose of development of the standard and for commercial implementations of the standard. At the Maceió meeting in November ’96 MPEG reached sufficient confidence in the stability of the standard under development, and produced the Working Drafts (WDs). Much work has been done since, resulting in the production of Committee Drafts (CD), Final Committee Drafts (FCD) and finally Final Drafts of International Standard (FDIS), in Atlantic City, October 1998.
The WDs already had the structure and form of a standard but they were kept internal to MPEG for revision. Starting from the Sevilla meeting in February ’97, MPEG decided to publish the WDs to seek first comments from industry. The CD underwent a formal ballot by national Bodies and the same happened to the FCD. This applies to all parts of the standard, although ‘Conformance’ is scheduled for later completion than the other parts.
Ballots by NBs are usually accompanied by technical comments. These ballots were considered at the March ’98 meeting in Tokyo for CD and in Atlantic City (October 1998) for the FCD. This process entailed making changes. Following the Atlantic City meeting, the standard was sent out for a final ballot, where NBs could only cast a yes/no ballot, without comments, within two months. After that, the FDIS became International Standard (IS) and was sent to the ISO Central Secretariat for publication. A similar process is followed for the amendments that form MPEG-4 Version 2. In December '99, at the Maui meeting, MPEG-4 Version 2 will acquire the status of FDIS.
Established in 1988, MPEG has grown to form an unusually large committee. Some 300 to 400 experts take part in MPEG meetings, and the number of people working on MPEG-related matters without attending meetings is even larger.
The wide scope of technologies considered by MPEG and the large body of available expertise, require an appropriate organization. Currently MPEG has the following subgroups:
| 1. Requirements | Develops requirements for the standards under development (currently, MPEG-4 and MPEG-7). |
| 1. Delivery | Develops standards for interfaces between MPEG-4 applications and peers or broadcast media, for the purpose of managing transport resources. |
| 2. Systems | Develops standards for the coding of the combination of individually coded audio, moving images and related information so that the combination can be used by any application. |
| 3. Video | Develops standards for coded representation of moving pictures of natural origin. |
| 4. Audio | Develops standards for coded representation of audio of natural origin. |
| 5. SNHC
(Synthetic- Natural Hybrid Coding) |
Develops standards for the integrated coded representation of audio and moving pictures of natural and synthetic origin. SNHC concentrates on the coding of synthetic data. |
| 6. Multimedia Description | Develops Structures for multimedia descriptions. This group only works for MPEG-7, |
| 7. Test | Develops methods for and the execution of subjective evaluation tests of the quality of coded audio and moving pictures, both individually and combined, to test the quality of moving pictures and audio produced by MPEG standards |
| 8. Implementation | Evaluates coding techniques so as to provide guidelines to other groups upon realistic boundaries of implementation parameters. |
| 9. Liaison | Handles relations with bodies external to MPEG. |
| 10. HoD
Heads of Delegation |
The group, consisting of the heads of all national delegations, acts in advisory capacity on matters of general nature. |
Work for MPEG takes place in two different instances. A large part of the technical work is done at MPEG meetings, usually lasting one full week. Members electronically submit contributions to the MPEG FTP site (several hundreds of them at every meeting). Delegates are then able to come to meetings well prepared without having to spend precious meeting time to study other delegates' contributions.
The meeting is structured in 3 Plenary meetings (on Monday morning, on Wednesday morning and on Friday afternoon) and in parallel subgroup meetings.
About 100 output documents are produced at every meeting; these capture the agreements reached. Documents of particular importance are:
Output documents are also stored on the MPEG FTP site. Access to input and output documents is restricted to MPEG members. At each meeting, however, some output documents are released for public use. These can be accessed from the home page: www.cselt.it/mpeg
Equally important is the work that is done by the ad-hoc groups in between two MPEG meetings. They work by e-mail under the guidance of a Chairman appointed at the Friday (closing) plenary meeting. In some exceptional cases, they may hold physical meetings. Ad-hoc groups produce recommendations that are reported at the first plenary of the MPEG week and function as valuable inputs for further deliberation during the meeting.
| AAC | Advanced Audio Coding |
| AAL | ATM Adaptation Layer |
| Access Unit | A logical sub-structure of an Elementary Stream to facilitate random access or bitstream manipulation |
| ACE | Advanced Coding Efficiency (Profile) |
| Alpha plane | Image component providing transparency information ??? (Video) |
| Amd | Amendment |
| AOI | Area Of Interest |
| API | Application Programming Interface |
| ARTS | Advanced Real-time Simple (Profile) |
| ATM | Asynchronous Transfer Mode |
| BAP | Body Animation Parameters |
| BAP | Body Animation Parameter |
| BDP | Body Definition Parameters |
| BDP | Body Definition Parameter |
| BIFS | Binary Format for Scene description |
| BSAC | Bit-Sliced Arithmetic Coding |
| CD | Committee Draft |
| CE | Core Experiment |
| CELP | Code Excited Linear Prediction |
| CIF | Common Intermediate Format |
| CNG | Comfort Noise Generator |
| DAI | DMIF-Application Interface |
| DCT | Discrete Cosine Transform |
| DMIF | Delivery Multimedia Integration Framework |
| DNI | DMIF Network Interface |
| DRC | Dynamic Resolution Conversion |
| DS | DMIF signaling |
| EP | Error Protection |
| ER | Error Resilient |
| ES | Elementary Stream: A sequence of data that originates from a single producer in the transmitting MPEG-4 Terminal and terminates at a single recipient, e.g. an media object or a Control Entity in the receiving MPEG-4 Terminal. It flows through one FlexMux Channel. |
| FAP | Facial Animation Parameters |
| FBA | Facial and Body Animation |
| FCD | Final Committee Draft |
| FDIS | Final Draft International Standard |
| FDP | Facial Definition Parameters |
| FlexMux stream | A sequence of FlexMux packets associated to one or more FlexMux Channels flowing through one TransMux Channel |
| FlexMux tool | A Flexible (Content) Multiplex tool |
| FTTC | Fiber To The Curb |
| GMC | Global Motion Compensation |
| GSTN | General Switched Telephone Network |
| HCR | Huffman Codeword Reordering |
| HFC | Hybrid Fiber Coax |
| HILN | Harmonic Individual Line and Noise |
| HTTP | HyperText Transfer Protocol |
| HVXC | Harmonic Vector Excitation Coding |
| IP | Internet Protocol |
| IPI | Intellectual Property Identification |
| IPMP | Intellectual Property Management and Protection |
| IPR | Intellectual Property Rights |
| IS | International Standard |
| ISDN | Integrated Service Digital Network |
| LAR | Logarithmic Area Ratio |
| LATM | Low-overhead MPEG-4 Audio Transport Multiplex: |
| LC | Low Complexity |
| LOAS | Low Overhead Audio Stream |
| LOD | Level Of Detail |
| LPC | Linear Predictive Coding |
| LSP | Line Spectral Pairs |
| LTP | Long Term Prediction |
| MCU | Multipoint Control Unit |
| Mdat | media data atoms |
| Mesh | A graphical construct consisting of connected surface elements to describe the geometry/shape of a visual object. |
| MIDI | Musical Instrument Digital Interface |
| MPEG | Moving Pictures Experts Group |
| MPEG-J | Framework for MPEG Java API’s |
| MSB | Most Significant Bits |
| NB | National Body |
| OCI | Object Content Information |
| OD | Object descriptor |
| OD | Object Descriptor |
| PDA | Personal Digital Assistant |
| PSNR | Peak Signal to Noise Ratio |
| QCIF | Quarter Common Intermediate Format |
| QoS | Quality of Service |
| Rendering | The process of generating pixels for display |
| RTP | Real Time Transport Protocol |
| RTSP | Real Time Streaming Protocol |
| RVLC | Reversible Variable Length Coding |
| SA-DCT | shape-adaptive DCT |
| SID | Silence Insertion Descriptor |
| SL | Sync(hronization) layer |
| SL | Sync Layer |
| SNHC | Synthetic- Natural Hybrid Coding |
| SNR | Signal to Noise Ratio |
| Sprite | A static sprite is a - possibly large - still image, describing panoramic background. |
| SRM | Session Resource Manager |
| T/F coder | Time/Frequency Coder |
| TCP | Transmission Control Protocol |
| TransMux | generic abstraction for any transport multiplex scheme |
| TTS | Text-to-speech |
| UDP | User Datagram Protocol |
| UEP | Unequal Error Protection |
| UMTS | Universal Mobile Telecommunication System |
| VCB | Virtual CodeBook |
| Viseme | Facial expression associated to a specific phoneme |
| VLBV | Very Low Bitrate Video |
| VM | Verification Model |
| VOP | Video Object Plane |
| VRML | Virtual Reality Modeling Language |
| WD | Working Draft |
| WWW | World Wide Web |
