1、W 4855232 0533q33 794 D 51 RECOMMENDATION ITU-R BR. 1356 USER REQUIREMENTS FOR APPLICATION OF COMPRESSION IN TELEVISION PRODUCTION (Questions ITU-R 238/11 and ITU-R 239/11) (1 998) The ITU Radiocommunication Assembly, considering a that new disk based storage supports are expected to penetrate all a
2、reas of television production, namely non-linear editing, on-air playout and short- to medium-tem archives; b) that this technology offers significant gains in terms of vastly improved operating flexibility, production flow and station automation and is therefore highly attractive for the up-grading
3、 of existing studios and the design of completely new studio installations; c new interconnecting networks which allow interactive and multi-user operation mandates however a significant reduction of bit rate of the video signal subjected to the above processes; d) data storage as well as different
4、file formats and networking protocols for signal interchange have been already introduced to the market which jeopardize interoperation between individual equipment and remote studios of different manufacture; e be particularly important and urgent and will be beneficial to broadcasters, including t
5、hose in developing countries, as it is demonstrated by contributions received from ITU-D, the WU and some administrations, that the economic and time efficient use of hierarchical storage technology together with that a number of mutually incompatible bit-rate reduction schemes to achieve economic t
6、hat studies of compression schemes for television data storage and archival are deemed to recommends 1 This implies availability to all interested parties on a fair and equitable basis of the intellectual property necessary to implement those standards. Availability in the marketplace of chip sets a
7、nd/or algorithms for software encoding and decoding may give users confidence in the adoption of particular compression methods; 2 uniquely defined television production application in order to maximize compatibility and interoperability ; that compression algorithms and transport schemes should be
8、based on Open Standards. that the number of compression methods and parameters should be minimized for each * The scope of this Recommendation is limited to 525-line and 625-line interlaced systems and video compression schemes using bit rates of 50 Mbit/s or less (excluding audio) and a coding reso
9、lution of one TV-frame or longer. A general tutorial description of compression algorithms suitable for use within television production as well as some application examples are given in Appendix A to this Recommendation. M:BRSGDPUBLI98SG 1 1 BRTEXTS-E 13 56.DOC D 4855212 0533434 b20 52 3 that compl
10、iance testing methods should be available for those building equipment to standards for algorithms and transport schemes and for users purchasing and installing equipment to those standards. Standards bodies should adopt standards for compliance testing methods to support both manufacturer and user
11、needs; 4 that a single compression scheme used with different compression parameters throughout the chain should be decodable by a single decoder; 5 that the development of a common (“agile”) decoder is desirable to support the use of more than one compression family; 6 that integration of video com
12、pression into more complex systems must be via standardized interfaces. Translating through Recommendation ITU-R BT.60 1, i.e., decoding and re-encoding, should be the default method of concatenating video signals compressed using different techniques andor parameters; 7 the serial digital interface
13、 (SDI) as embodied in ITU-R BT.656; 8 allow predictable levels of performance to be achieved in the implementation of specific applications; 9 that bit streams carrying compressed signals should be designed so that they can be formatted and packaged for transport over as many types of communications
14、 circuits and networks as possible; 10 that appropriate channel coding methods and error protection be employed where necessary; 11 that compression systems should be designed so that, in normal operation, signal timing relationships (e.g. audiohideo lip sync) and synchronization presented at encode
15、r inputs are reproduced at decoder outputs; 12 durations that are practical for specific television production applications; 13 carried through the compression system, although not necessarily compressed with the video. Provision should be made to carry selected parts of metadata through a transpare
16、nt path synchronously with the video and audio data; 14 that the compression scheme chosen for devices that mimic VTRs should allow for the reproduction of pictures in shuttle mode for identiing content and of pictures in jog and slow motion modes for selecting edit points; 15 that network interface
17、s and storage devices should provide for both Variable Bit Rate (VBR) and Constant Bit Rate (CBR) options; 16 that storage devices should allow recording and playing back of television programme streams and files as data rather than decoding to base-band for recording and re-encoding upon playback;
18、that the compression scheme chosen should not preclude the use of infrastructures based on that issues related to interoperability must be further explored and standards developed to that signal delays through compression processing (encoding/decoding) must be limited to that provision should be mad
19、e for selected analogue vertical interval information to be M:BRSGDPUBLI98SG 1 lBRTEXTS-E 1356.DOC 53 17 that the compression strategy chosen for standard television should be extensible to high definition applications to allow for commonality in the transitional phase. NOTE - Specialized terms freq
20、uently used in the context of innovative television production are listed in Recommendation ITU-R BR. 1357. APPENDIX A Report on the use of compression in television production Introduction Digital compression for video represents the core enabling technology for innovative television programme prod
21、uction of the future since it permits the storage of programme material in servers that make access to that material virtually instant and allows simultaneous use by multiple users. These features all have the effect of improving workflow efficiency and reducing the cost of producing, post-producing
22、 and repurposing of television programmes. The report identifies a number of parameters which define the basic performance of different compression schemes and enlarges on their impact on picture quality and post-production headroom within a variety of applications normally encountered in distribute
23、d television production. PART A - COMPRESSION ISSUES 1 Digital compression for video 1.1 Image quality Selection of compression system parameters has a significant impact on overall image quality. These compression parameter choices must be optimized to preserve image quality while at the same time
24、fitting the image data into the available bandwidth or storage space. Different combinations of compression parameters may be best for different specific applications. Compression system parameters which should be considered include: the underlying coding methods, the coding sampling structure, pre-
25、processing, data rates and the group of pictures (GOP) structure used. In choosing compression system parameters, interaction between parameter choices must also be considered. Finally, special operational issues such as editing the bit stream or splicing new content into an incoming bit stream shou
26、ld be considered. 1.1.1 Coding method The coding method is the most fundamental of compression choices. There are four main compression methods used in the television production and distribution chain: MPEG-2 Main Profile at Main Level (MPML), MPEG-2 4:2:2 Profile at Main Level (4:2:2ML), Motion M:B
27、RSGDPUBLI98SG 1 lBRTEXTS-E1356.DOC - m 4855212 053343b 4T3 m 54 JPEG, and DV. All of these coding methods are based on the Discrete Cosine Transform (DCT). All of these coding methods use normalization and quantization of the transform coefficients, followed by variable length coding. MPEG includes
28、motion estimation and compensation in its tool kit of techniques. This allows improved coding efficiency, with some cost penalty in memory and processing latency. Motion JPEG and DV are both frame-bound, thereby minimising coding cost, but these frame-bound coding methods do not take advantage of th
29、e coding efficiency of inter-frame motion estimation and compensation. MPEG and DV both allow motion adaptive processing in conjunction with intra-fiame processing. 1.1.2 Sampling structure MPEG, Motion PEG, and DV can all be used with the 4:2:2 pixel matrix of ITU-R BT.601, MPEG and Motion JPEG can
30、 both be used with other pixel matrices, multiple frame rates, and either interlace or progressive scan. Note that the 4:2:2 matrix is sub-sampled from the original full-bandwidth (4:4:4) signal. The pixel matrix can be further sub-sampled to reduce the signal data, with 4:2:2 sampling normally used
31、 for interchange between systems. 4:2:2 systems, such as the MPEG-2 4:2:2 Profile, 4:2:2 Motion JPEG, and the DV 4:2:2 system (which operates at 50 Mbitss), all use half the number of colour difference samples per line compared to the number used in the luminance channel. 4:2:2 provides half the hor
32、izontal bandwidth in the colour difference channels compared to the luminance bandwidth and full vertical bandwidth. 4: 1 : 1 systems such as DV 525 use one quarter the number of colour difference samples per line compared to the number used in the luminance channel. 4: 1 : 1 reduces the colour diff
33、erence horizontal bandwidth to one quarter that of the luminance channel while maintaining full vertical bandwidth. The filters used to achieve the 4: 1 : 1 sub-sampled horizontal bandwidths, like other horizontal filters, generally have a flat frequency response within their pass-bands thereby enab
34、ling translation to and from 4:2:2 with no further degradation beyond that of 4: 1 : 1 sub-sampling. 4:2:0 systems such as DV 625 and MPEG-2 Main Profile, use half the number of colour difference samples horizontally and half the number of colour difference samples vertically compared to the number
35、used in the luminance channel. 4:2:0 therefore retains the same colour difference horizontal bandwidth as 4:2:2 (Le., half that of the luminance channel) but reduces the colour difference vertical bandwidth to half that of the luminance channel. 4:2:0 coding, however, generally does not provide flat
36、 frequency response within its vertical pass-band, thereby precluding transparent translation to the other coding forms. Consequently, systems that use 4:2:0 sampling with intermediate processing generally will not retain the full 4:2:0 bandwidth of the prior coding. Care must be exercised in select
37、ing compression sampling structures where different compression coding techniques will be concatenated. In general, intermixing different sub-sampled structures impacts picture quality, so cascading of these structures should be minimized. For example, while 4: 1 : 1 or 4:2:0 signals will have their
38、 original quality maintained through subsequent 4:2:2 processing (analogous to “bumping up” of tape formats), cascading 4: 1 : 1 and 4:2:0 will generally yield less than 4: 1 :O performance. Note that the proposed SMFTE D-7 format, although based on DV coding, will use 4: 1 : 1 sampling for both the
39、 525- and 625-line systems. 1 M:BRSGDPUBLI98SGl lBRTEXTS-E13 56.DOC 4855232 0533437 33T m 55 1.1.3 Compression pre-processing Video compression systems have inherent limitations in their ability to compress images into finite bandwidth or storage space. The compression systems rely on removal of red
40、undancy in the images, so when the images are very complex (having very little redundancy), the ability to fit into the available data space may be exceeded, leading to compression artefacts in the picture. In these cases, it may be preferable to reduce the complexity of the image through other meth
41、ods before compression processing. These methods are called pre-processing, and include filtering and noise reduction. When noise is present in the input signal, the compression system must expend some bits encoding the noise, leaving fewer bits for encoding the desired image. When either motion det
42、ection or motion estimation and compensation is used, noise can reduce the accuracy of the motion processing, which in turn reduces coding efficiency. Even in compression systems which do not use motion estimation and compensation, noise adds substantial energy in high frequency components of the DC
43、T which might otherwise be zero. This not only wastes bits on extraneous DCT components, but degrades run length coding efficiency as well. Compression system specifications generally define only the compression functions within equipment, but do not specify the pre-processing before the compression
44、 function. An exception is the shuffling which is an inherent part of the DV system, and is not to be confused with the shuffling used for error management in digital recorders. Since most pre-processing, such as filtering or noise reduction, is not always required, the pre-processing parameters may
45、 be selected depending on the nature of the images and the capabilities of the compression system. These choices can be pre-set or can be adaptive. 1.1.4 Data rate The MPEG-2 4:2:2 Profile at Main Level defines data rates up to 50 Mbits/s. Motion JPEG 4:2:2 equipment typically operates at data rates
46、 up to 50 Mbit DV operates only with constant bit rate. In practice, even those systems commonly believed to be constant data rate have data rate variation, but over shorter periods of time. Another way to characterize the compression systems is to compare constant quality with constant data rate sy
47、stems. Constant quality systems attempt to maintain a uniform picture quality by adjusting coded data rate, typically within the constraint of a maximum data rate. Since simpler images are easier to code, they are coded at lower data rates. This results in more efficient compression of simpler image
48、s and can be a significant advantage in storage systems and in non-real-time transfer of images. Constant quality operation is useful for disk recording and some tape recording systems such as tape streamers. Constant quality vs. constant data rate Constant data rate systems attempt to maintain a co
49、nstant average data rate at the output of the compression encoder. This will result in higher quality with simpler images and lower quality with more complex images. In addition to maintaining a constant average data rate, some constant data rate systems also maintain the data rate constant over a GOP. Constant data rate compression is useful for videotape recording and for fixed data rate transmission paths, such as common carrier services. Constant data rate processing will, of course, be characterized by a target data rate. Variable data rate processing
