WO2001013646A1

WO2001013646A1 - Method and apparatus for coding

Info

Publication number: WO2001013646A1
Application number: PCT/GB2000/003205
Authority: WO
Inventors: John Andrew Fletcher; Paul Burfield
Original assignee: British Broadcasting Corporation
Priority date: 1999-08-17
Filing date: 2000-08-17
Publication date: 2001-02-22
Also published as: GB2353426A; GB9919449D0

Abstract

A coder and method is disclosed for solving the novel problem of providing multiple coded versions of an input signal efficiently. The signals are preferably coded using common analysis circuitry, and preferably combined into a single multi-programme transport stream. The coder may be modular, or composed of discrete components.

Description

METHOD AND APPARATUS FOR CODING

The present invention relates to a coder for producing a compressed version of an input video or audio signal. The invention is particularly applicable to production of a compressed MPEG video signal.

Pursuant to the invention, it has been appreciated that there are applications where it may be desirable to produce more than one compression encoded version of an input video (or audio) signal each having a different bit rate. For example, a television channel such as BBC 1 may be coded at 5.032Mbit/s for digital terrestrial transmission, 8.655Mbit/s for feed to national regions and at a variable rate in the range of 2-8Mbit/s for feeding to a statistically multiplexed digital satellite broadcast link. Another application contemplated pursuant to the invention is a situation where it is desired, for example, to view or listen to a signal over a variety of different quality communication links.

Such applications can be realised in a straightforward manner using conventional commercially available equipment, by providing a plurality of coders, each one configured to provide the appropriate bit rate output.

Pursuant to the invention, we have appreciated that, although the coding is, of course, different for each of the different bit rates, a large portion of the processing, in particular the analysis required for coding, may be duplicated amongst the coders.

By way of background, EP-A-559 982 discloses a system for coding a signal for storage on a digital video recorder in which a multi-level coder produces a coded output partitioned into a low level coded constant bit rate stream, which can be used to provide some output during high speed playback, and an additional data stream used to correct vector quantising errors in the low-level version to enable more acceptable quality normal playback to be produced. A specialised decoding method improves the low level version based on the error correcting data stream when available. The use of this multi-level coding in which the lowest level is a low quality coded version of the input signal and the other ievel(s) data streams are not complete but merely "fill in" gaps in the low level is disclosed to provide efficient coding for the purposes of storing on videotape. This method for use in a video recorder in which it may be possible to employ both proprietary compression and decoding is of course not suitable for producing broadcastable output to be decoded by conventional decoder and is not intended to provide more than one complete coded output.

According to a first aspect, the invention provides apparatus for coding an input signal comprising an analyser for analysing the input signal to provide analysis information for use in coding; a first compression encoder for providing a first compression encoded signal based on the input signal and said analysis information and having a first bit rate; and a second compression encoder for providing a second compression encoded signal based on the input signal and said analysis information and having a second bit rate.

In this way, multiple (independently decodable) signals (decodable bitstreams) having different bit rates may be efficiently produced in a single coder (from the same input signal) avoiding excessive duplication of circuitry. A further advantage is that signals of different bit rates may have a greater degree of correspondence since identical analysis can be performed. The invention can be extended to produce third and further coded outputs.

In a second aspect, the invention provides a modular coder for producing multiple compression encoded outputs comprising: means for receiving an input signal; an analyser for analysing the input signal to produce analysis information for use in coding; and means for supplying the input signal and the analysis information to a plurality of compression encoders.

In this way, as many coders as are required can be added to a modular system. The compression encoders may have independently configurable output bit rates. The means for supplying may include a motherboard or distribution outlet with a plurality of connection points or sockets, preferably buffered, into which cards or plugs may be inserted. Power and other data may also be supplied. It will be appreciated that the modular coder is just a component of a complete coding system; to form a working system, one or more compression encoders must also be provided. The invention extends also to a system, and to a compression encoder adapted for use in such a modular coding system. The input signal may be pre- processed prior to supplying to the compression encoders.

In one preferred embodiment, the coder is provided with a housing capable of housing a plurality of compression encoders; this can enable a compact multi-output device to be provided.

In another preferred embodiment, the analysis unit is arranged to supply a plurality of separate compression encoders; this may enable greater freedom to choose different compression encoders to be used for different bit rates. In such a case, the information is most preferably supplied with the video signal using a defined format. A most preferred format is the MOLE™ format, as prescribed by the British Broadcasting Corporation and Snell and Wilcox Ltd.

Preferably, the analysis information includes a list of motion vectors, or information indicative of motion vectors. Determination of motion vectors is a complex process and so providing this information can simplify the compression encoders.

It will be appreciated that, because analysis information is provided to the coders, the coders can be relatively simple coders, relying predominantly on the analysis information for coding decisions.

In a preferred application, the outputs of the coders are combined into a single multi-programme transport stream. In this way, a single stream can be supplied to multiple separate pieces of equipment which can each select an appropriate bit rate signal from the stream; this may facilitate interconnection of equipment, particularly in a complex multi-programme, multi-output distribution environment, and may simplify re-configuration. The combination of different bit-rate versions of a single signal into a single stream may be provided independently.

In a related third aspect, the invention provides a compression encoded multi-programme transport stream comprising a plurality of compression encoded versions of an input signal, the versions having mutually different bit rates. Preferably, the versions are coded using at least some common coding decisions provided by a common analysis unit, most preferably coded using motion vectors estimated by a common motion estimator.

In a first method aspect, the invention provides a method of coding a signal comprising: receiving an input signal; analysing the input signal to produce analysis information for use in coding; and supplying the signal and the analysis information to a plurality of compression encoders each coupled to receive the input signal and the analysis information and configured to produce a respective compression encoded output signal having a specified bit rate.

In a second method aspect, the invention provides a method of coding an input signal comprising compression encoding the signal to produce a plurality of coded signals, each having a respective bit rate, and combining the coded signals into a multi-programme transport stream.

The signal may be an audio or video signal, or most preferably an audio/video signal (which term is meant to include a signal (or signals) capable of carrying audio and video). The coders are preferably arranged to perform MPEG coding, particularly MPEG-2 coding, although derivatives and successors of the standard may be employed, and coders coding according to mutually different standards may still be supplied with at least some of the same analysis information.

An embodiment of the invention will now be described, by way of example, with reference to the accompanying drawings in which:-

Figure 1 shows a typical set up in which multiple outputs at different bit rates are combined into transport streams for distribution;

Figure 2A shows the arrangement of a conventional coder which can be modified according to the invention;

Figure 2B shows a modified version of the coder of Fig. 2A, arranged to produce 3 outputs in a single transport stream; Figure 3 shows an alternative arrangement in which an analysis unit transmits coding information to separate coders. Referring first to Fig. 1 , a typical application of the invention is shown. This is applied to MPEG-2 coding of audio/video, for television transmission. In this arrangement, it is desired to receive multiple channel inputs and to provide feeds, at varying bit rates to different transmission outputs, namely a digital terrestrial output, a sustaining network feed and a statistically multiplexed satellite output. This could be achieved by multiple independent commercial coders, but at significant cost. If conventional coders are used, it is nonetheless desirable to multiplex the separate outputs of a single programme onto a common transport stream, to simplify distribution to the various outputs. Much more preferably, as will be described, it is desirable to derive the outputs making use of common analysis circuitry.

Referring to Fig. 2A, the schematic arrangement of an exemplary MPEG-2 coder is shown. Most commercially available coders have a similar basic architecture. To modify the coder according to the invention, additional

"dim" coders DCD1 , DCD2, DCD3 are fitted in place of the original coder DCD, connected to receive the signal and analysis data produced by preceding components. The analysis data supplied preferably includes at least motion vectors as these are complex to derive. References to motion vectors being supplied are intended to include motion vector precursors comprising information from which motion vectors may be readily derived but not formatted as motion vectors.

It is desirable also for the analyser to determine a GOP (group of pictures) structure and to supply information concerning Picture Type (l,P,B), Prediction Type (intra frame, forward/backward frame/field), DCT Type

(frame or field) and Differential Quantisation. Although the main quantisation scale will be determined by each compression encoder based on the desired bit rate, a differential value is added to this based on picture content, and this will be common among the encoders. If all of the above information is supplied, the compression encoder can determine the quantisation scale simply from knowledge of the bit rate required and coding should require minimal further analysis.

The above information applies particularly to MPEG-2 coding of video; other equivalent information may be substituted for other forms of coding and for other data. As will be appreciated, the general principle is to put as much of the analysis that can be shared into a common analysis unit as possible.

In applications where the outputs are of very similar bit rates, it is preferable to carry out most of the analysis in common, and only to vary the quantisation scale between the compression encoders to achieve the desired bit rate. However, where greater differences in bit rate are required, further differences between the encoders may be desirable.

In some applications, it may be desirable for the compression encoders to perform some independent analysis, and to refine or modify the analysis information supplied by the analysis information. This may be based on prediction selection based on a locally decoded picture (as the eventually decoded picture will be generated by applying predictions to decoded pictures). Alternatively, it may be desirable to modify the GOP structure. For example, if a very high bit rate output is required (e.g. 50 Mbit/s), it may be desired to increase the proportion of l-frames, to facilitate editing.

Similarly, if very low bit rate is desired, the GOP length may be increased. The signal resolution may vary between the encoders.

The analyser may be configure to include alternatives in the analysis information so that compression encoders may select appropriate data without having to perform independent analysis. For example, the analyser may determine GOP structures appropriate to a high bit rate and to a low bit rate and include information concerning both in the analysis information.

The bitstream formatting and rate control is most easily implemented based on the existing coder by employing individual components for each coder.

It will be appreciated that the compound coder requires only a small amount of additional circuitry to produce multiple outputs compared to the single output coder, and considerably less than the conventional solution of providing multiple independent coders. It will of course be apparent that the compound coder can produce independent (complete) output bitstreams which can be independently decoded.

The general principle of the multiple-output coder is that coding decisions are determined by a common analysis front-end. These decisions are then conveyed, together with the corresponding picture data, to a number of simple coders which perform rate-control and quantisation. The coder outputs may be multiplexed into a single transport stream. Alternatively, independent outputs may be provided if desired.

The coding decisions and picture data can be conveyed by various means. In the above example, the signals are conveyed at a low level within hardware.

In a modular coder, it is preferable to make the relevant signals available at a plurality of slots or sockets into which coders may be plugged as required. The signals are preferably supplied together with appropriate power and any other necessary signals for a given implementation, and preferably connection points for outputs. Such a coder may advantageously be built into a housing with space for multiple coders, and coder cards may be added as required.

Another possible implementation is to use multiple discrete coders, and to signal analysis data to each coder. The analysis data may be supplied along a dedicated signal bus or connector. However, a preferred method is to use the so-called MOLE™ where coding decisions are "buried" alongside the video on a standard digital video interface, e.g. the Serial Digital Interface

(SDI). Although the MOLE™ was intended for use when re-coding video, we have appreciated that it could also be used to control first-time encoding by multiple coders, thereby simplifying the coders. Although the MOLE™ is an advantageous method of burying data in a video signal, other methods of transporting data, particularly coding decision information, may be employed as required.

A MOLE™-assisted encoder designed to work only with a MOLE™ signal, i.e. not able to make its own coding decisions, would be comparatively simple and cheap to manufacture. It is in principle possible to remove the front-end of any suitable commercial coder and replace all the complex analysis circuitry with a MOLE™ interpreter to drive a "dim" coder, in much the same way as described above to insert further DCD coders into the coder of Fig. 2A. Figure 3 shows a configuration for a system employing multiple MOLE™ assisted coders.

Although the invention has been described above in the context of coding

MPEG video signals, the preferred application, it will be appreciated that the invention may be applied to coding of other signals where multiple coded forms are required, and where there is common analysis to be performed.

For example, in coding audio, for example MPEG Layer I, II, III, psychoacoustic analysis of the input signal is carried out as a first step. There is also usually a time to frequency transformation. Both of these steps are computationally intensive but are independent of the bit rate of the audio output. Thus, an MPEG audio coder with multiple outputs could advantageously have a common analyser performing psyco acoustic analysis and transformation, and separate encoders for each output supplied by the analyser.

Although the input signal will normally be supplied to each compression encoder, to enable maximum flexibility, this will normally have been buffered, so that it is co-timed with the analysis information (which may take some time to derive) and so the signal supplied will not be exactly the input signal. Where some pre-processing of the input signal is common to all coders, regardless of bit rate, for example a transformation, the analyser may include a pre-processor and the input signal supplied may be a pre- processed version of the input signal. References in the specification (which term includes the claims) to supplying the input signal to the compression encoders are intended to encompass supplying buffered, delayed or pre-processed versions of the input signal.

Whilst the coding described above is primarily to produce similar outputs, differing only somewhat in bit-rate, it will be appreciated that the outputs may differ greatly in nature, for example one may be a broadcast quality output of at least about 5 Mbit/s or 8 Mbit/s and another may be highly compressed low-bandwidth, for example for archiving merely to identify scenes or for transmission over a telephone line or the Internet. As shown, the outputs may include fixed and/or variable bit rate outputs. Modifications of detail may be made, particularly in the case of coding of signals other than video. The appended abstract is incorporated herein.

Claims

1 . Apparatus for coding an input signal comprising: an analyser for analysing the input signal to provide analysis information for use in coding; a first compression encoder for providing a first compression encoded signal based on the input signal and said analysis information and having a first bit rate; and a second compression encoder for providing a second compression encoded signal based on the input signal and said analysis information and having a second bit rate.

2. A modular coder for producing multiple compression encoded outputs comprising: means for receiving an input signal; an analyser for analysing the input signal to produce analysis information for use in coding; and means for supplying the input signal and the analysis information to a plurality of compression encoders having independently configurable output bit rates.

3. A modular coder according to Claim 2, further comprising at least one compression encoder.

4. A compression encoder for use with the modular coder of Claim 2 arranged to receive analysis information for use in coding from an external analyser.

5. Apparatus according to any of Claims 2 to 4, wherein the means for supplying comprises a plurality of connection points or sockets into which cards or plugs of compression encoders can be inserted.

6. Apparatus according to Claim 5, wherein the modular coder is arranged to supply power to the compression encoders.

7. Apparatus according to any of Claims 2 to 6, further comprising a housing capable of housing a plurality of compression encoders.

8. Apparatus according to any of Claims 2 to 6 arranged to supply a plurality of discrete compression encoders.

9. Apparatus according to Claim 8, wherein the analysis information is supplied with the video signal in a defined format.

1 0. Apparatus according to any preceding claim, wherein the analysis information includes a list of motion vectors.

1 1 . Apparatus according to any preceding claim, wherein the analysis information includes Picture Type and/or Prediction Type.

1 2. Apparatus according to any preceding claim, wherein the analysis information includes DCT Type.

1 3. Apparatus according to any preceding claim, wherein the analysis information includes Differential Quantisation.

14. Apparatus according to any preceding claim, wherein the analysis information, together with information specifying a desired bit rate, enables the compression encoders to determine quantisation scale.

1 5. Apparatus according to any preceding claim, wherein at least a portion of the analysis information comprises alternatives so that different compression encoders can select different alternatives, appropriate to their respective bit rates.

1 6. Apparatus according to any preceding claim, wherein at least one compression encoder is capable of analysing the input signal and independently deriving analysis data to replace a portion of said analysis information.

1 7. Apparatus according to any of Claims 1 to 1 5, wherein the or each compression encoder is arranged to code based on the analysis information as supplied by the analyser.

1 8. Apparatus according to any preceding claim wherein the analyser includes means for pre-processing the input signal prior to supplying it to the or each compression encoder.

1 9. Apparatus according to any preceding claim, further comprising means for combining the outputs of the coders into a single multi- programme transport stream.

20. A compression encoded multi-programme transport stream comprising a plurality of compression encoded versions of an input signal, the versions having mutually different bit rates.

21 . A signal according to Claim 20, wherein the versions are coded using at least some common coding decisions provided by a common analysis unit.

22. Apparatus according to any preceding claim wherein the signal is an audio/video signal.

23. Apparatus according to any preceding claim arranged to perform MPEG coding.

24. A method of coding a signal comprising: receiving an input signal; analysing the input signal to produce analysis information for use in coding; and supplying the input signal and the analysis information to a plurality of compression encoders each coupled to receive the input signal and the analysis information and configured to produce a respective compression encoded output signal having a specified bit rate.

25. A method according to Claim 24, further comprising combining the coded signals into a multi-programme transport stream.

26. A method of coding an input signal comprising compression encoding the input signal to produce a plurality of coded signals, each having a respective bit rate, and combining the coded signals into a multi-programme transport stream.

27. A method according to Claim 24, 25 or 26 wherein the signal is an audio/video signal.

28. Apparatus or a method according to any preceding claim wherein at least one output signal is a broadcast quality signal, preferably having a bit rate of at least about 5 Mbit/s.