INTERNATIONAL ORGANISATION FOR STANDARDISATION
ORGANISATION INTERNATIONALE DE NORMALISATION
ISO/IEC JTC1/SC29/WG11
CODING OF MOVING PICTURES AND AUDIO

ISO/IEC JTC1/SC29/WG11N1417

mpeg96
November 1996 / Maceió

Highlights

During the 37th MPEG meeting, from 18 through 22 November in Maceió, Brazil, hosted by ABNT (the Brazilian national standards body), the first draft versions of the MPEG-4 Audio (ISO/IEC 14496-3) and Video (ISO/IEC 14496-2) parts standard were issued. This means that an enormous amount of technical work in the 5 MPEG meetings held this year, has resulted in first, relatively stable descriptions of the final MPEG-4 standard. The Systems part (ISO/IEC 14496-1) has already been at the state of Working Draft since November 1995, and during the Maceió meeting, version 2.0 was issued.

The subjective multichannel tests carried out for MPEG-2 ‘Advanced Audio Coding’ AAC, previously known as ‘Non-Backwards Compatible Audio’) showed that the quality delivered at 320 kbits/s was of ‘EBU indistinguishable quality’. The AAC coding scheme is the first to achieve this quality under the rigorous test conditions employed. The low complexity mode of AAC also performed excellently at this bitrate, missing the ‘EBU indistinguishable quality’ by a hair’s breadth.

Building upon the success of the recently completed MPEG-2 Digital Storage Media Command and Control (DSM-CC) international standard, MPEG announced the next phase of DSM-CC as part of the MPEG-4 work effort. This phase will be known as the DSM-CC Multimedia Integration Framework (DMIF). DMIF will expand upon the current DSM-CC standard (ISO/IEC 13818-6) to enable the convergence of interactive, broadcast and conversational multimedia into one specification which will be applicable to settops, desktops and mobile stations.

MPEG now has a clear picture of the goal of the next MPEG-standard, ‘MPEG-7’. The name of the MPEG-7 work item is ‘Multimedia Content Description Interface’, that will allow fast and efficient searches for audiovisual content - anywhere in the world. It will provide the tools to build the multimedia analogy of the very popular text-based search-engines that are currently available on the WorldWide Web. More information can be found on the MPEG homepage: http://www.cselt.stet.it/mpeg. Also other information about MPEG work can be found there.

Details

This section lists details about the work, arranged according to the subgroups in MPEG.

Audio

MPEG-2 Audio

MPEG has reviewed the results of subjective listening tests on the MPEG-2 Non-backwards compatible (NBC) multichannel audio coding algorithms which were carried out by the BBC and NHK between 16 September and 11 October 1996.

The tests evaluated the following codecs

MPEG-2 NBC at 256 kbit/s

MPEG-2 NBC at 320 kbit/s

MPEG-2 NBC low-complexity version at 320 kbit/s

1995 version of MPEG-2 Layer II at 640 kbit/s in a backwards compatible mode.

The test procedure and environments complied with ITU-R Recommendation BS-1116. Listener reliability and test procedure checks were included and a detailed statistical analysis of the results was performed. The results showed good performance for all of the codecs. The MPEG-2 NBC codec at 320 kbit/s generally performed better than the other codecs although, overall, the MPEG-2 NBC low complexity version delivered only marginally worse performance.

These tests evaluated only a subset of the full set of NBC tools and only assessed the multichannel performance. It is worth emphasizing that these tests were conducted according to the most rigorous of test methods. These deliberately emphasize the differences between stimuli. Comparisons to other test results using less rigorous methodologies should not be made.

The Committee Draft of MPEG-2 NBC (13818-7) has been successfully progressed to the status of Draft International Standard.

MPEG decided that as a result of the success of the NBC coding experiments and the rapid progress towards ‘International’ status of the proposed Standard, that the time was now right to choose a permanent name for this part. The Standard will in future be known by the name MPEG-2 Advanced Audio Coding (AAC).

MPEG-4 Audio

Good progress on the development of the audio verification model (VM: a model of the standard that can be implemented to carry out so-called ‘Core-Experiments’) for MPEG-4 has been made. New proposals for core experiments have been assessed and, where additional functionalities are being offered, the proposals have already been married into the VM. Both the encoder and decoder audio VMs will be based on software modules to speed up the process of collaborative developments. The decoder modules will be free of copyright for the purpose of MPEG-4 optimization.

The Working Draft of the proposed MPEG-4 Audio Standard (14496-3) has been prepared, as have detailed contributions to the MPEG-4 specification of requirements. Additional input on these requirements is being sought such that they can be finalized at the February meeting of MPEG.

Video

In the MPEG-4 Video Group, numerous technical contributions have been reviewed, and the most promising proposals have been accepted as a part of the draft standard. These new elements of the MPEG-4 standard improve considerably the motion-compensated prediction. (As in many video coding schemes, in consecutive frames, only the differences between two frames are transmitted. That difference is calculated by compensating the motion between the two frames.) The resilience against transmission errors has been improved by means of desynchronizing markers. Some new techniques have been adopted for coding the shape of non-rectangular video objects, which is a unique feature of MPEG-4. These tools considerably improve the efficiency of coding this shape.

Before the meeting in Brazil, a large number of ‘Core Experiments’ have been carried out, where experts try to improve the currently used technology. In the time between this meeting in Brazil and the next meeting in February 1997, only the most promising of these core experiments will be continued. In this way, the standardisation converges by focusing its activity on the points with the highest potential of enhancement of the standard.

The meeting has released the first Working Draft of MPEG-4 video, a first draft of the final standard. The final standard shall have the official state of "Committee Draft" in November 1997. The final standard specifies the bitstream syntax and semantics and the decoder. For its common experiments, the Video Group continues developing a reference encoder, called "Verification Model" (VM). For that VM, an improved rate control has been accepted.

Systems

The Systems Group released the Systems Working Draft 2.0 and Systems Verification Model 2.0.

The Working Draft contains two main normative parts. The first one describes precise specifications of the syntax and semantics for 2D composition. The future release of this section will also address 3D composition. The second one describes the multiplex and synchronization of audio-visual information. This specification, originally based on the ITU H.223/A standard evolved toward a two layer model. The advantage of a two layer model is that it allows flexible adaptation to different networks, an important requirement for MPEG-4. The higher layer, the so-called "content multiplex" allows a preliminary grouping of multimedia information streams that share common characteristics, reducing overhead and delay.

A new version of the Verification Model was released, that will allow Systems experiments. The current experiments target, in particular, integration of multiplex, events modeling, concurrent implementation and syntax configuration. The Systems sub-group chose to support, like the Audio, SNHC and Video Groups, a collaborative development of a complete Systems software Verification Model.

SNHC (Coding of synthetic audiovisual content)

The SNHC Group is further focusing on visual technology to allow virtual meetings and on combining moving video with text/graphics overlays. This technology is useful in news broadcasts, advertising and business information. On the audio side of the work, the focal point of the work is the integration of ‘Text-to-Speech’ and synthetic audio, very low-bit-rate speech coding (in the order of a few hundreds kilobits/second) These techniques can be used for facial animation, ‘story telling on demand’, and virtual environments. It is worth noting that the very low bitrate audio coding employed here can be used for both natural and computer-generated speech.

Test

The Test Group has selected the methods and defined the conditions for the July ’97 MPEG-4 tests, that have two main goals: a) to verify the performance of audio and video technology, looking at the functionalities which MPEG-4 will provide, but also from an application point of view, and b) to evaluate new audio and video proposals against the VM and existing standards.

Companies represented in MPEG showed particular interest to evaluate the performance of the video Verification Model in error-prone environments, and the audiovisual quality of MPEG-4 in a setting of real-time communication. These tests will be carried out under conditions that are as realistic as possible. When available, hardware will be used for these tests.

DSM-CC (Digital Storage Media - Command and Control)

As was stated in the ‘highlights’ section, the DSM-CC Group has started to work on the ‘DSM-CC Multimedia Integration Framework (DMIF), as a part of the MPEG-4 work effort. This work encompasses:

The concept of a common end-to-end "session" interface across multiple network provider implementations

Integration with multiple network technologies, such as ATM and the Internet

Integration with specific object domains, such as CORBA and JAVA

Multiple devices participating as peers within the same session

Fully symmetric consumer and producer operations within a single device

The scheduling and real-time switching/multiplexing of bit streams

The highly desirable integration feature of DMIF will provide the tools to inter-operate across diverse access and core networks such as Hybrid Fiber Coax (HFC), Local Area Network (LAN), Asynchronous Transfer Mode (ATM), and the Internet. The DMIF "session" concept associates resources using a template which contains the resource parameters necessary for inter-operability and accounting.

Typical applications which will benefit from this initiative are interactive multimedia retrieval; video telephony; audio, video and graphics conferencing; and networked multiplayer games. These applications can all be operated from the same device.

Since the purpose of DMIF is to integrate across different platforms and networks, MPEG welcomes the participation of all standards groups and industry organizations in a cooperative effort to contribute to the DMIF initiative.

Implementation Studies Group

The Implementation Studies Group reviewed changes to the video decoder verification model introduced since the previous meeting. The group was of the opinion that the new techniques in the area of ‘padding’ (a very frequently used technique when coding objects with arbitrary shape) will significantly reduce the computation complexity and will aid the further analysis to assist in the decision making process of the standardization effort. Additionally, the group discussed system architecture approaches that will provide means to exploit various levels of parallel processing which will help to ensure that systems can be affordably constructed and yet still be able to deliver the level of performance required by MPEG-4.

Background information

The next meeting of MPEG will take place in Sevilla, Spain, from 17 to 21 February 1997. As this is an extra meeting, MPEG will also next year go beyond the regular three meetings.

For further information on this press release and MPEG work in general, please contact:

Dr. Leonardo Chiariglione, (Convenor of mpeg)

CSELT

Via G. Reiss Romoli, 274

10148 Torino, ITALY

Tel.: +39 11 228 6120; Fax: +39 11 228 6299

Email: leonardo.chiariglione@cselt.stet.it

or refer to the MPEG homepage:

http://www.cselt.stet.it/mpeg

and to other pages containing information on MPEG-4 elements:

http://www-elec.enst.fr/msdl

http://www.es.com/mpeg4-snhc