1993 - von John R Forrest National Transcommunications Ltd (NTL) UK

Digital videocompression is not a new technique. Even as early as the 1970s, work was being carried out in NTL's laboratories (then the IBA Research and Development Department) on compression codecs for a system to provide inter-studio links and to distribute the programmes of the planned UK fourth UHF television channel to the transmitter sites.

A single frame of a PAL television picture has approximately 700 x 570 picture elements (400,000 pixels). In a colour picture it is necessary to specify the levels of three colours (red, green, blue) at each pixel, so 1.2 million numbers are required. The human eye finds it difficult to discriminate levels which differ by less than about 1%, so 8-bit computer numbers are quite adequate to describe the colour levels at each pixel. A single frame therefore contains 1.2 Mbytes of digital information, which is approximately the storage capacity of a 3h" floppy disk.

The fundamental process of videocompression is the same for both still and moving pictures. The picture or frame is segmented into square blocks of pixels (for example, 8x8 pixels). Each block is analysed in terms of regular patterns in the levels of brightness and colours. It is also possible to exploit certain characteristics of the human eye: the eye is not very sensitive to fine detail and it is less sensitive to variations in colour than in brightness. Less accuracy and hence less digital information is therefore required in the colour signals and in the fine brightness detail. The net result is that compression ratios of up to 50:1 may be achieved with very little degradation for the majority of still pictures.

The compression processing of full motion video signals is more demanding, since successive frames of the PAL picture arrive at intervals of 1/25th of a second. However, since in general each frame differs little from the previous one, the essential part of the process is the subtraction of one frame from the next in sequence, leaving only the differences to be transmitted. The differences are due to motion of some kind in the picture and since motion involves areas of significant numbers of pixels together and is in most cases regular, it is possible to simplify the information needed even further. This "motion compensation" processing is a key part of the process. The more recent advanced processing schemes also allow for much better processing to be achieved by predicting new frames, interpolating between these and then carrying out forwards and backwards comparison to refine the predictions.

Work on these processing schemes, or algorithms, has centred on an international collaborative group known as MPEG (Moving Pictures Experts Group) . This group also has the role of providing standards input to ISO and has been greatly aided by input from the European collaborative EUREKA project known as VADIS. The first phase of standardisation, known as MPEG 1, was achieved in 1991 and describes the internationally-agreed compression scheme suitable for medium resolution video in CD-I type applications. The next phase, suitable for broadcast applications, known as MPEG 2, has almost completed the standardisation process; the vision format was agreed earlier this year and the complete specification including the multiplex is anticipated before the end of 1993.

The MPEG standards have become accepted internationally by all those with major involvement in the field of videocompression. This has been one of the most successful development and standardisation programmes in recent times, contrasting strongly with chaotic situations in DBS and HDTV transmission standards.

The data rate for full motion video signals of studio quality is in the region of 200 Mbit/s (approximately - 1 Mbyte x 8 bits x 25 frames/sec). Since the mid 1980s such digitised video could be carried with very little degradation on the standard 14 0 Mbit/s circuits of the major telecommunications carriers. The expense of the codecs and the wideband digital circuits, however, restricted severely the amount of signals conveyed in this form. Nevertheless the increasing use of digital equipment in studios and the immunity to degradation of digital signals resulted in pressure to find ways to reduce these costs. An important step was the development and standardisation by the European Telecommunications Standards Institute (ETSI) of a 34 Mbit/s compression codec suitable for studio quality broadcasting applications. This opened prospects of a wide market and led to low-cost codecs. An important step was taken by the European Broadcasting Union (EBU) with the announcement in 1992 that the Eurovision network would become a 34 Mbit/s digital network.

In 1993, with codecs commercially available from a number of suppliers, major implementations of 34 Mbit/s links have occurred, for example, for Channel 4 programme distribution by British Telecom and for regional television services and the linking of cable television and telephony networks by NTL.

While 34 Mbit/s studio quality compression systems have been in the process of being brought into service, developments on more highly compressed video suitable for the less demanding requirements of home television have been rapid. In 1989 it was confidently predicted that compression to 34 Mbit/s was the limit for broadcast studio applications, but that in the home environment 17 Mbit/s would be the limit for demanding material (sports, horse racing) and 8 M/bits the limit for the typical video movie.

As shown in Fig. 1, development systems for studio applications are now being demonstrated at 8 Mbit/s and fully acceptable quality for video movies is being achieved in the latest system from NTL and Scientific Atlanta at bit rates as low as 2 Mbit/s. A reluctance to predict the situation in 1995 is understandable!

There has even been investigation of transmitting signals of video movie quality along standard telephone lines to subscriber premises. A system known as asymmetric digital subscriber line (ADSL) , proposed by AT&T in the USA, would use a 1.5 Mbit/s compressed signal. The concept, as shown in Fig. 2, is that of having a local video switch which, by access to a video library, provides a single demanded service (the latest popular movie, for example) along the subscriber line. The system has been demonstrated in the laboratory, but it is not known how it would cope in practice with the very variable quality of domestic and local telephone wiring! Despite the apparent simplicity of the system, it is likely to involve as large or likely greater financial investment than installing cable systems to the home.

The progress that has been made with digital compression in the last two years is impressive. Much is due to microprocessor development allowing faster, more complex processing, to better techniques for motion compensation, and the availability of low-cost, rapid-access digital memory to store the information during processing. The differentiation between systems now lie in the way which systems cope with motion in the picture, the speed with which they adapt to new scenes (such as a cut between cameras) and the way they recover from errors (such as interference to the signal). An important matter commercially is the system complexity and the partitioning of complexity between encoding and decoding. This will have a strong bearing in the end cost of decoders for the domestic market.

Digital compression is applicable to all transmission media and it is important that there should be as much commonality of technique and equipment as possible between systems for the various media.

It has been conventional to use a 36 MHz transponder for each television programme channel with the current uncompressed analog signals. It is now possible to put between 2 and 8 programme channels on such a transponder, depending on the quality of video signal that is required. Such systems clearly have a major impact on satellite programme delivery costs per channel, but also allow expansion in the number of channels in areas where transponder availability is at a premium. The former case applies particularly in businesses such as news feeds and distance learning, the latter in areas such as South America. Argentina has already implemented a multi-channel digital compression feed system to cable heads. In the USA, Hughes and Thomson have announced a major multi-channel direct-to-home satellite service, called "Direct TV", for inauguration in early 1994.

Many currently-installed cable systems have capacity for about 50 conventional uncompressed analog channels. This has been adequate until now for live programming, but the newer fibre-optic cable systems are offering up to several hundred channels to allow inclusion of video movie-on-demand and interactive services.

Typically, it is possible over a current 50 channel system to provide some 150 channels of compressed video with little extra investment in the cable transmission network - a major part of the original capital cost of the system. New decoder units are required in subscribers' homes, but the cost of these may be amortised more directly over the increased revenues from the new services.

Both TCI and HBO in the USA, who operate major cable networks, have indicated plans to offer video-on-demand and interactive services in the near future.

The current radio frequency spectrum allocations in each country allow a very limited number of television channels to be transmitted. It is typically less than six in most European countries and in the UK has remained at four nationwide channels for some years.

Digitally compressed signals have two interesting advantages in the terrestrial application. First, up to four such video services could be transmitted, depending on the picture quality required, in the frequency spectrum currently occupied by one programme channel. Secondly, the digital signals can be arranged to co-exist and not interfere with the existing PAL transmissions. This allows a transition period, such as that from black-and-white 405-line transmissions to 625-line colour transmissions to occur.

NTL has developed, under contract to the Independent Television Commission (ITC), such a digitally-compressed UHF television transmission system, comparable with wide-screen programme material, known as SPECTRE. It was recently demonstrated using new low-power digital transmitters at the transmitting sites in Devon of the ITV and Channel 4 networks. A key question for regulatory bodies such as the ITC is as to how this new capability should be used. It could deliver new programme services, movie-on-demand services, or high definition television (HDTV) services. It is the latter route which appears to be under adoption by the FCC in the USA, but in Europe there is not a conviction that a market yet exists for HDTV services.

The issue of standards in areas of technology, particularly where the market volume of products is high, is frequently vexed and emotive.

Although doubtless many debates will still run, the key aspects of videocompression standardisation have occurred with the successful work of MPEG. The most crucial area in digital compression for mass-market applications is the realisation of low cost domestic decoders. This requires the processing circuits to be implemented in a standard form on a small number of very large scale integrated circuit (VLSI) silicon chips, available from more than one manufacturing source. The way ahead to this is now clear. The early announcements from TCI and HBO, major cable system operators, that their equipment purchases would have to conform to the MPEG standard was particularly helpful.

The situation in the digital sound compression encoding is less satisfactory. Stemming from the use of digitally-compressed sound in many applications (film, CD, tape) for a number of years, there are a variety of incompatible proprietary formats around (eg. APT, Dolby AC3) in addition to the MPEG audio standard. A resolution of these to a single internationally-agreed standard is likely to be very difficult and the cost-effective solution may be to implement most of these on a single silicon chip and allow the one relevant to the particular programme service to be automatically selected. Processing of audio signals is relatively simple and the overhead of such a multiple implementation is not high.

Encryption and conditional access is the area where standardisation is most difficult. Digital signals are inherently easy to put into secure encrypted form, but it is likely that programme providers will wish to retain a number of proprietary conditional access systems.

It might be expected that there would be a requirement for standardising the transmission bit rate for digitally compressed video applications. The CCITT in the telecommunications field has standardised bit rates over a wide range of applications. The range from 64 kbit/s to 2Mbit/s has been particularly important for telephony, data and now videoconference. Likewise, the range from 34 Mbit/s to 14 0 Mbit/s and above has been well used for trunk telecommunications. Apart from the data rate used by compact disks (4.4Mbit/s), the range 2 Mbit/s to 34 Mbit/s has received less prominence.

However, a defined hierarchy of bit-rate standards for video-compressed signals may not now be necessary. What is interesting about the advanced digital compression video encoders now available, such as the system available from Scientific Atlanta/NTL, is that they are fully flexible in data rate between 8 and 2 Mbit/s, easily selected by the user according to the type of programme material and application. The decoder in turn automatically adjusts to the incoming data rate.

Most of the major broadcasters in the world are now undertaking trials of videocompression equipment. Current encoder prices are in the range $100,000, depending on functionality. Decoders are in the range $3,000. Since relatively few encoders are required, the price is not very critical at this stage. The next key milestone will be the availability of low-cost decoders at a typical price of some hundreds of dollars, for which very substantial investment in VLSI chips will be required. These may be expected in early 1994, consistent with the timescales of various major ventures such as the "Direct TV" project in the USA.

Little has been heard from Japan about the evolution of digital compression systems, but a very significant announcement was that from Sony and Matsushita recently concerning their joint development of a digital VCR which will undoubtedly use compression.

The VCR is important in two respects: it is needed in the home to record digital broadcasts directly and avoid the complexities and distortions of conversion to PAL, SECAM or NTSC; also the VCR is the means of playback of pre-recorded video material. It seems likely that compressed digital recordings of movies will become available at about the same time as the first digital broadcasts and will have a significant effect in extending the penetration of the wide-screen format.
Digital video-compression can be expected to be in relatively widespread use in broadcasting, including the first steps into the domestic market within two years. In due course there will be compression encoding and decoding technology on a few silicon chips suitable for camcorders and a host of other imaging applications.

The most significant aspect of this evolution, however, is its common impact across what have previously been separate media. The MPEG standardisation process is likely to mean that there will be interoperability between different storage media (semiconductor, tape, optical disk) and between broadcast and computer equipment. This will lead to the video equivalent of the printed magazine publishing and distribution industry. Videocompression is, in short, the major vehicle for multi-media convergence between the film, broadcast and computer worlds.