EUROPEAN EVOLUTION T OWARD HD M. Croll1, P.Bohler2, L. Haglund 3 1
BBC R&D, UK, 2NRK, Norway, 3SVT, Sweden
ABSTRACT The paper will describe work that has been co-ordinated by the EBU Television Quality Evolution Group to understand what can be done to deliver the best possible pictures via digital television delivery systems to the new, large flat-screen displays. This work has involved an examination, by European Broadcasters of how they can optimise the delivery of standard definition television and how they could adopt a common format to deliver HD. The format that is recommended is 1280x720P/50 The paper describes the results of extensive investigations aimed at getting the very best digital delivery systems for Europe. This work has taken due account of the current commitment of European countries to their standard – definition digital television services and the efficiency needed to deliver improved services in the strong competition for terrestrial services throughout Europe. INTRODUCTION The high cost and early commitment to unsuccessful HD in Europe (HD MAC and Eureka 95) leaves a legacy of doubt that any HD proposals could ever be successful in Europe. But now HD is fast becoming a viable consumer proposition because: •
Large (30-40 inch) flat-LCD screens are now becoming available at affordable prices
•
HD broadcasts are successful in Japan and the US
•
Multi-megapixel still picture cameras as well as HD camcorders are available as consumer products
•
HD DVD’s are available, together with domestic equipment to record them.
In addition, the advance of technology and the specific European Broadcasting situation makes it inappropriate for Europeans to follow the precise route to HD established in Japan and the US because: •
New flat-screen displays are progressive – interlace does not fit comfortably with new compression and display technology
•
Compression technology now delivers higher efficiencies than those achieved with MPEG2
•
Congestion of Europe’s airwaves demands the highest efficiency to deliver the widest diversity of higher resolution television services
New, large screens are becoming available at affordable prices. These are able to display the new HD services in the US and Japan to advantage. But European standard definition services look inferior on such displays.
A group of European Broadcasters has formed the EBU B/TQE to consider Television Quality Evolution. They have recognised the developing consumer HD environment as both an opportunity and a potential threat, especially in maintaining the high value of their programmes. This paper describes measures that can be taken to get the very best out of the current standard-definition MPEG-2 based services that European Broadcasters have committed to. It then presents proposals for a format for HD in Europe (the 1280x720P/50 format) to deliver the benefits of HD to the widest possible European audience. GETTING THE BEST OUT OF CURRENT STANDARD DEFINITION TELEVISION Current European digital services, with the exception of “Euro 1080”, deliver 576 lines in a 625-line format via MPEG-2 in either 16:9 or 4:3 aspect ratio. The data-rate chosen for these services is generally much less than the average used with DVDs and in many cases the bitrate chosen is so low as to introduce visible artefacts. Overall, the quality of such digital television services, when they are displayed on a large screen, is determined by a large number of factors, not just the bit-rate. Broadcasters will always want to broadcast at the lowest possible bit-rate to deliver the largest number of channels with the richest data interactive services; broadcast channel capacity is a valuable commodity. To achieve the objectives of individual broadcasters, optimising their delivery chain to best deliver to large flat-screens implies: •
Delivering to the encoder pictures which are easiest for the coder to process having regard to the programme maker’s intentions and the impact the pictures will have.
•
Using encoder arrangements that will result in the minimum artefacts being introduced into difficult to code sequences
Care and attention to all aspects of production and post-production and minimum use of composite sources or links helps in avoiding the worst effects. A detailed set of guidelines for best practice can be found in (1). Collections of the more important ones to remember are listed below. In following the signal flow from acquisition to play-out the list may be as follows: Quality Aspects and Capture •
Programme makers need to understand the different types of scene content and the way they behave in compression.
•
Quality is more than technical parameter values alone and is also influenced by the professional skills in shot framing and scene composition of the cameraman and editor in the production process.
•
When deciding on the adequacy of technical parameters it is important to remember that in addition to the basic picture quality you will need headroom to be included in these parameter values although they may not be immediately visible. This headroom represent the safety factor against impairments as the signal undergoes further processing.
•
The ‘quality factors’ also include elements such as colour balance, colorimetry, lighting and the effects of contrast and noise.
•
Avoid low level lighting and high-gain settings (introduces noise) and be careful with the use of ‘aperture correction ‘ and/or ‘contour/detail’. This is particularly important in
low-cost type cameras (i.e. DV camcorders). Use the best possible lens to minimise the need for these corrections Processing •
Uncompressed ITU-R BT.601 production will produce the best quality to be delivered to the encoder. In today’s server and- file based production infrastructure this may well be impractical. Nevertheless, 4:2:2 sampling structures should be used throughout the entire production and contribution processes.
•
In the production chain multiple decoding and recoding must be avoided. Compressed video (video files) should be carried throughout production in its ‘native’ compressed form (as when it emerged from the camera).
•
For mainstream TV production do not compress to bit rates below 50Mbps as explained in (2).
•
For file transfer, the MXF file format should be used as it provides standardised methods for mapping native compressed (and uncompressed) video and audio.
•
If, in the overall chain, multiple codec’s cascading cannot be avoided, then at least similar encoding and decoding devices should be used to minimise quality loss (3). Do not under any circumstance introduce signals with a PAL or SECAM history into the digital signal chain.
Primary Distribution •
MPEG-2MP@ML encoding should be used for transmission and it is very important that encoders having a very high performance are used.
•
The highest possible bit-rate should be used together with statistical multiplexing if more than two programs are being distributed in the same stream.
•
The performance of MPEG-2 encoder may vary significantly from manufacturer to manufacturer, users should evaluate available encoders either with their own tests or based on reports of the experience of others. As a rule of thumb, across the broadcast chain, the same type of MPEG-2 encoders provides better overall quality than a mixture of types and makes.
Having considered the advice given above there is yet another way of obtaining an improved quality in the decoded picture. Adopting HD equipment in the production chain will increase the ‘headroom’ in the signal quite considerably. Down converting to the SD level prior to encoding for transmission will give an enhanced picture quality compared to the quality obtained even with a “pure” ITU-R BT.601 production chain. Coder implementations like ‘two pass’ encoders have also improved since MPEG-2 coders were first introduced. These encoders inconnection with statistical multiplexing can increase the quality efficiency by up to 20% and there are new proposals that have been found to benefit some critical sequences (4). Figure 1
COMPARING STANDARD DEFINITION AND HD DELIVERED IMAGES Overall, whilst there is scope for optimisation of the standard definition image delivered to large flat-screens, the image nearly always falls short of the equivalent HD delivered image. Figure 1 shows a comparison of SD delivered with current MPEG-2 compared with HD delivered using a more advanced coding method. The figure shows small sections Picture Quality SD of a complex moving sequence. Figure 2 shows how SD and HD have been compared by visitors to a home fair in Sweden. In this case, very good quality standard definition DVDs were compared with HD Digital VHS on adjacent screens. The results were that the SD picture quality was generally “fair” whilst the HD picture quality was judged as “excellent”. A RECOMMENDATION FOR HD BROADCAST FORMAT FOR EUROPE
No answer
Bad Poor
Fair
Good Excellent
Picture Quality - HD 80.0%
76.9%
The EBU B/TQE group have been 70.0% investigating the range of formats that are 60.0% in use for delivering HD at 60Hz in the US 50.0% and Japan and the European (50Hz) 40.0% 30.0% equivalents that could be adopted. The 17.2% 20.0% Group, and the EBU Technical Committee 10.0% 4.8% (its parent group in the EBU) recognised 0.4% 0.4% 0.1% 0.0% that there was no need to link delivery No Bad Poor Fair Good Excellent format with production format. A wide answer range of production formats is needed to support the rich variety of programmes. Figure 2: Comparison of assessments of SD Individual programmes use combinations and HD of film and a range of video and HD formats that are chosen to suit the workflow of the production. This very large range of formats is best converted to a common format before delivery. In this way, professional conversion equipment can be used to offer an optimised signal to the coder and conversion of image formats within consumer equipment is minimised. The EBU Technical Committee agreed in April 2004 to recommend to EBU members the use of a progressively scanned delivery (or transmission) format for higher quality digital television broadcasting by its members. They noted that there are two candidate progressively scanned formats included in SMPTE specifications that are both capable of conveying full motion or film motion. These are 720P50 and 1080P/50. The EBU Project Group B/TQE has been evaluating these two options. They hoped to establish which of them could be proposed to EBU Members as the recommended format for broadcasting higher quality television. But they also considered the individual requirements of broadcasters who might want to adopt HD broadcasting at different times and in different situations. Consequently, they also reviewed whether more than one format should be recommended. The evidence before the group to-date leads to the conclusion that a 720P/50 format should
be recommended. Some of the background to this recommendation is given in the following sections. The need for a “progressive” delivery format A progressive delivery format is consistent with the “progressive” way that new displays work and it is easier to compress with new compression technology. It can be output directly from current picture sources or generated from interlace TV sources using very high quality conversion equipment. New LCD, plasma and non-CRT based projection technology are very different from the CRT technology they replace (5). The new screens are “progressive” in that the entire picture is refreshed at one time. The CRT technology uses “interlace” so that only half of the information of a picture is refreshed at a time. It is very easy to convert a progressive delivered image to an interlaced form (to maintain compatibility with the older CRT displays) but its much more difficult to convert an interlaced image to progressive form to suit it to the new displays. Indeed, the EBU group spent some time assessing de-interlacers common in the domestic display environment and found they generally contributed substantial impairment and limited the final quality of an HD delivered image. However, professional conversion equipment of very good performance has been developed and good de-interlacers are available from a range of manufacturers. Tests with H264 and Windows media 9 (WM9) compression systems have established that they compress progressive images better than they compress interlace images. The bit-rate required to deliver a good quality 720P image has been found to be less than that required to deliver 1080i by an amount greater than the difference between the uncompressed datarates. It is noted that new compression systems sometimes introduce picture impairments when they interpret interlace “twitter” effects as movement within a picture. This limits the overall picture quality assessment of an HD image when it is delivered as an interlaced image and is undesirable. Current picture sources are fundamentally “progressive”. The CCD, at the heart of each camera, converts the optical image into electrical form with charges from all the rows of CCD elements transferred into a storage device at the same instant. “Interlaced” or “progressive” images are formed when the signal is “read” out of the chip, indeed the interlaced signal is formed by discarding information. This is done to reduce the analogue bandwidth of the signal – this is not an issue with digital delivery. Much electronic graphic programme material is generated in progressive form to avoid the twitter or flicker of fine detail. Conversion from interlace to progressive is complex and requires access to a number of successive fields – (6) - (with a consequent delay to the picture). Some equipment has been constructed that uses motion-compensation to improve the conversion process. In general this conversion is best done using professional equipment. The EBU B/TQE group found that the general quality achieved with such conversions in domestic equipment was very poor and contributed quality impairments that were clearly visible with a wide range of picture material. From this, we conclude that conversion from interlace to progressive should not be carried out at the receiver.
The screen sizes likely to be purchased in future The display manufacturers and 40% broadcasters have conducted extensive surveys to establish the size of flat screen display that 30% consumers are likely to purchase. Figure 3 shows results from a survey that was conducted by SVT in Stockholm. It asked visitors to a Home Fair to consider the large flat screen as “furniture” and consider what sizes would be desirable in their homes. The results were very similar to those quoted by other sources and confirmed that the majority of large flat screens are likely to be in the range 30 to 40 inch.
20% 10%
32”
42”
50”
60”
Larger sizes
Figure 3: Percentage of audience who would purchase different sizes of flat-screen TV screen
The Resolution needed to fill a 30-40 inch screen Tests have been conducted to establish how much detail is needed for different screen sizes in the home. In a first series of tests some 170 people measured the viewing distance if they were to introduce a large flat screen into their home. The results showed that the median viewing distance would be about 2.8m (7). This figure is remarkably similar to that measured 15yrs ago and does not seem to change much from country–to-country within Europe.
. Figure 4: TV Standard required to provide adequate horizontal resolution for 2.8m viewing distance
In a further series of tests, a range of still pictures were displayed on a monitor and observers asked to adjust the level of detail (using pre-filtered versions of the pictures). The monitor actually displayed small portions of the pictures with two versions alongside each over; one filtered and the other unfiltered. The results are shown in figure 4 expressed as the number of observers requiring a particular TV standard to provide adequate horizontal resolution at different screen sizes. From Figure 4 it is clear that the 720P50 standard would be best able to deliver to new large flat screens between 28 inches and 50 inches. Optimising the broadcast services in Europe Transmission bandwidth is a very scarce commodity in Europe. Each country makes extensive use of terrestrial, satellite and cable to deliver a rich diversity of broadcast services. In some areas of Europe, 3 or 4 countries have boundaries that are very close and each community expects local services. This sets very high standards for the efficiency of services and has caused most countries to use digital television to introduce multi-channel services where each programme channel has a low bit-rate allocated to it. In some European countries the further development of digital services is seen as a chance to improve the technical quality of what is broadcast. One example of this is Sweden. SVT, Sweden, has been at the forefront of the debate for a higher definition delivery system for European (public service) broadcasters. SVT went for multicasting of SD several years ago, as did several other early digital providers in Europe, This enabled them to distribute a large number of services very cost-effectively. However, SVT finds the balance between quality and quantity of programs (hence services) a difficult thing. This is because Sweden has few TV-households and the total amount of money that viewers are willing to pay for content is not increasing. It is hard to predict exactly what the viewing audience will equip themselves with, but they are interested in flat panels as furniture. Unlike the US, it is expected that European governments will not get involved in the broadcasting of HD. However, SVT is steadily producing more and more programs in HD so that it is prepared for broadcasting HD at the appropriate time. Reaching consensus among European broadcasters about interlaced or progressive scan for future higher resolution transmissions has been important for SVT as it affects the choice of optimum production equipment. THE WAY FORWARD To continue to deliver a multi-channel proposition, but to allow improved quality images to be delivered, demands a very high standard of efficiency for a new, higher definition TV service. This is why the EBU B/TQE group are confident that they can propose the 1280x720P50 standard as the single standard to deliver better images to larger flat screens. In making this proposal, the group expect the service to be optimised so that no significant picture impairment is visible on flat screens up to 50inches. Whilst it is likely that 10% of large screens in homes could be larger than this, it is not expected that the image format would deliver a particularly poor picture on a 60inch screen. 60% of the viewers would be expected to notice the difference between this format and a higher format (delivered via DVD – say) but, overall, the deficiencies would be no more distracting than the lack of resolution of SD pictures are on a 32 screen now.
This choice of format allows new compression systems to be applied efficiently so there is an expectation that compression artefacts will not be significant. In fact, since such new services would be delivered specifically to put good pictures on new large flat screens, it is hoped that sufficient care will be taken in coding the pictures and that a sufficient bit-rate will be allocated so that pictures are substantially free from compression artefacts. In this way, even the SD viewer or the observer who sits a greater distance from the large screen would derive benefit from such a service. The next stages will be to better assess the standard of service that can be achieved and to obtain some measure of the number of premium services that could be delivered via terrestrial, satellite and cable systems. Clearly the character of new compression systems will need a new study of the optimum multiplexing arrangements. One new feature that is expected to shape new services to large screen displays is the use of non-realtime delivery. Clearly the receiving equipment for new services will be very different from that currently used for standard definition. This opens the chance to specify advanced features, like non-realtime, as a way of achieving a multi-channel service for premium programmes at peak hours. Including this within the service proposition and specifying that new HD receivers should be able to operate like this would also maximise the data allocated to the delivery of each programme. Future work must also address the lack of 1280x720P50 support. This has hampered demonstrations and needs to be solved before European broadcasters can commit to commencement of a service. CONCLUSIONS The work reported here has convinced members of the EBU B/TQE Group that a single standard could be selected for Europe that would overcome difficulties European Broadcasters have in delivering good quality image to new large flat-screen displays. The Group unanimously decided for 720P50. Further, the Group decided it was appropriate to canvas other broadcasters and members of the broadcast and consumer industry for their reactions in the hope the EBU would be able to confirm the recommendation before September 2004. REFERENCES 1. EBU Technical Information I 39 – 2004 (Maximising the quality of conventional quality broadcasting in the flat panel environment). 2. EBU Technical Statement D84 - 1999 (Use of 50Mbps MPEG compression in television programme production). 3. EBU Report BPN 034 (Reference chain for the production of a TV programme; Appendix C). 4. EBU Report BPN 037 (Final report on statistical multiplexing). 5. Salmon, R.A. 2004. The Changing world of TV displays – CRTs challenged by flat panel displays. BBC White Paper WHP 089 6. Thomas, G.A. 1996 A comparison of motion compensated interlace – to progressive conversion methods. BBC R&D Report 1996/9 7. Tanton, N.E. 2004, Results of a survey on television viewing distance. BBC R&D White Paper WHP 090
