1、EG 38-2001 SMPTE ENGINEERING GUIDELINE MPEG-2 Operating Range Applications Contents 1 Scope and application 2 Normative references 3 Introduction to operating points 4 Professional production system data flow 5 MPEG application examples 6 Key issues by application area Annex A Related documents Anne
2、x B Bibliography 1 Scope and application 1.1 Scope The aim of this document is to provide practical guide- lines to users of MPEG-2 in studio and in other pro- f essional appl cat ions. This guideline provides a system overview, detailing the elements to be considered when choosing an MPEG-2 operati
3、ng range. This guideline describes how the structure and parameters defined in SMPTE RP 213 may be con- figured to meet a selected operating point. This ic achieved by giving specific, but representative, imple- mentation examples planned or in use around the world. Examples are included from both i
4、ntraframe and tem- poral predictive coded MPEG-2 for both 422 profile at main level and high level. 1.2 Application The picture quality obtained when decoding an MPEG encoded signal is largely dependent upon the combi- nation of bit rate and GoP structure. The results of tests made by the SMPTE and
5、EBU are shown in figure 1 and are as reported in the EBU/SMPTE task Page 1 of 12 pages force report (for full details of the test procedures please refer to the report). Figure 1 illustrates that for a given sequence of images, the same image quality can be obtained by various combinations of bit ra
6、te and GoP structure. The curves shown in figure 1 should be taken as an indication of first-generation performance within the wide span of MPEG-2 encoding options only. Taking signal differences as a measure of picture quality only allows coarse evaluation of actual quality perform- ance. The varia
7、nce of encoding parameters allowed in MPEGQ encoding structures to achieve the desired flexibility will require subjective testing of each individ- ual encoder design to determine the actual quality performance at a given data rate and GoP structure. Thearrows indicate possible members of the MPEG-2
8、 4:2:26 ML compression family envisaged for main- stream broadcasting, news and sports, and for contri- bution, as implemented in current industrial designs. Since this figure was produced, for example, temporal predictive coding efficiency has improved allowing lower data rates than those given in
9、figure 1. A specific operating point for mainstream production has been defined at 50 Mb/s and this is described in SMPTE 356M. 2 Normative references The following standards contain provisions which, through reference in this text, constitute provisions of this guideline. At the time of publication
10、, the editions indicated were valid. All standards are subject to revision, and parties to agreements based on this guideline are encouraged to investigate the possibility of applying the most recent edition of the standards indicated below. Copyright O 2001 by THE SOCIETY OF MOTION PICTURE AND TELE
11、VISION ENGINEERS 595 W. Hartxlale Ave., White Plains, NY 10607 (914) 761-1 100 Approved August 30,2001 - EG 38-2001 Bit-rate required to achieve a YN of 4 dB asa function I I I I IP IB IBPB N=l5 M=3 Figure 1 - Basic characteristics of compression for news and sports - MPEG-2 4:2:2PML SMPTE 312M-2001
12、, Television - Splice Points for MPEG-2 Transport Streams SMPTE 327M-2000, Television - MPEG-2 Video Recoding Data Set SMPTE 356M-2001, Television - Type D-i O Stream Specifications - MPEG-2 4:2:2P ML for 525160 and 625/50 SMPTE RP 202-2000, Video Alignment for MPEGP Coding SMPTE RP 213-2001, MPEG-2
13、 Operating Ranges 3 Introduction to operating points SMPTE RP 213 defines the MPEG-2 operating ranges. These ranges constrain characteristics of the MPEG-2 4:2:2 profile to ensure bitstream interchange in the professional environment. The operating ranges are subsets of ISO/MPEG pro- files and level
14、s. SMPTE RP 213defines two operating ranges for standard-definition television and three operating ranges for high-definition television. It also defines a hierarchical relationship among the ranges. This guideline analyzes the production system data flow, describes the operating range applicable fo
15、r each main production stage, and gives examples of operating points. For each stage in the production system data flow, the guideline discusses considera- tions that influence selection of an operating range and within it an operating point or points. Specific considerations include quality of vide
16、o and audio, efficiency, technology, and existing equipment and services. This guideline elucidates choice of operating points to enable bitstream interchange within professional appli- cations. It enhances the users ability to specify equip- ment and set application-specific parameters. The clauses
17、 that follow will inform and help guide the users to an optimum choice for their applications. Clause 4 describes the production system data flow and defines its stages. Page 2 of 12 pages Clause 5 gives, for each stage, example data rates and operating ranges. This is done separately for SDN and HD
18、TV. Further considerations for each stage follow. Clause 6 lists applications that use the production system described in clauses 4 and 5 and identifies issues that need to be considered when implementing a system. 3.1 Nomenclature The following definitions will be used in the clauses that follow: 3
19、.1.1 acquisition data rate: The (net or total) data rates used in field acquisition recording. 3.1 2 contribution data rate: The (net or total) data rates used to back-haul signals to a network center or station operating facility. 3.1.3 distribution data rate: The (net or total) data rates used to
20、pass network signals to regional or local stations. 3.1.4 editing data rate: The (net or total) data rates used for a reasonable number of post-production generations, not including the most demanding high-end post-production applications. 3.1.5 emission data rates: The (net ot total) data rates use
21、d in digital television broadcasting, either by terrestrial or satellite paths. 3.1.6 headroom: The additional data rate neces- sary to permit a naive transcode or naive com- pressed family conversion while retaining an acceptable picture quality. 3.1.7 intelligent compression family conversion: The
22、 conversion between the MPEG and non- MPEG (DV, JPEG, etc.) families of compression using nonessence data (such as bit-rate alloca- tion and quantization) in the recompression to minimize quality loss. 3.1.8 intelligent transcoding: T h e c o nv e rs i o n between two MPEG-2 formats using MPEG-2 rec
23、oding data set in the recompression to maxi- mize quality. The formats may differ in GoP structure, bit rate, frame rate, or frame size. EG 38-2001 3.1.9 naive compression family conversion: The conversion between the MPEG and non- MPEG families (DV, JPEG, etc.) of compression using only uncompresse
24、d video and audio in the recompress ion. 3.1.10 naive transcoding: The conversion be- tween two MPEG-2 formats using only uncom- pressed video and audio in the recompression. The formats may differ in GoP structure, bit rate, frame rate, or frame size. 3.1.11 net data rate: The data rate of the vide
25、o elementary stream. 3.1.12 on-air operations data rates: The (net or total) data rates used for local storage prior to master control in an on-air playout facility, and may be subsequently recoded for analog trans- mission or lower bit-rate digital transmission. 3.1.13 total data rate: The net data
26、 rate plus audio, data, and any system overhead. 3.1.14 transcoding: The process of converting one MPEG-2 format to another. 4 Professional production system data flow Figure 2 illustrates the various strategies available to broadcasters when designing a compressed video production flow process. In
27、this example, the split between long GoP and I-frame-only systems at 30 Mb/s reflects the point of approximately equal per- formance for long GoP systems at 30 Mb/s and l- frame-only systems at 50 Mb/s. Acquisition: The program may be captured in I-frame only or using long GoP MPEG-2 to provide high
28、er storage capacities (principally camcorders). Contribution: Where there is a need to electronically transmit the captured information over a network (sat- ellite/telco/wireless camera), the long GoP MPEG-2 format is more likely to be used as this provides for higher transmission efficiency. Source
29、 storage: Ideally the signal should be stored in its received format, that is I-frame-only for I-frame- only systems and long GoP for long GoP systems to maintain the highest possible quality. It is also possible to transcode the incoming feed to permit a standard Page 3 of 12 pages EG 38-2001 Q, ci
30、 c .- m 4 O .- A 50 40 230 20 10 O b Process Flow Transcode to Emlsslon Format Figure 2 - Compressed video process flow - Current practice - SD example native storage format. However, care must be taken to ensure that the quality is not degraded. Editing: Cuts only editing is simple to perform in th
31、e MPEG-2 domain in the case of l-frame-only VTRs or servers. Where more complex editing is required, the signal has to be processed at video baseband. This can be achieved using either I-frame or long GoP MPEG-2 provided that the system has sufficient head- room. Where this headroom does not exist,
32、it is nec- essary to use the MPEG-2 recoding parameters as defined in SMPTE 327M if MPEG-2 concatenation artifacts are to be avoided. Program storage: Program storage should also be made in the editing format or in the transmission format so that the number of transcoding stages is reduced to a mini
33、mum. The key options available to the program maker can therefore be summarized as: - An I-frarne-only or long GoP system with a suffi- ciently high data rate to allow multiple naive decod- inghecoding processes while maintaining the overall qual it y. - A long GoP system using lower data rates but
34、passing forward recoding information as described in SMPTE 327M to minimize MPEG-2 concatenation artifacts. 4.1 System issues To maintain quality throughout the production chain, the following issues should be taken into considera- tion: 1) The same GoP structure should ideally be main- tained throu
35、ghout the production chain in order to minimize the number of transcoding stages re- qui red. 2) The alignment of the MPEG-2 video should con- form to that defined in SMPTE RP 202. 3) For long GoP inputs or systems, recoding data should be used to minimize concatenation artifacts by feeding forward
36、previous encoding decisions as de- fined in SMPTE 327M. In practical systems and for economic reasons, it may be necessary to transcode between one form of Page 4 of 12 pages EG 38-2001 MPEGQ and another, for example, to pass through a Telco or satellite contribution circuit. 4.2 Transcoding issues
37、Multiple stages of compression encoding and decod- ing will introduce artifacts in the program content. Some conversions between different bit rates or com- pression structures will also introduce artifacts. Where practical, compressed program material should be transported and stored in the origina
38、l compression format to minimize the number of cascaded encoding and decoding stages in the production chain. How- ever, for many applications, it is necessaiy to mix sources derived from different MPEG compression parameters to create a final program. In these in- stances some form of transcoding w
39、ill be required. A method for reducing the number of transcoding stages is to define a set of encoding parameters to be applied as consistently as possible within an applica- tions area. To ensure the consistent operation of encoders between different manufacturers, it is nec- essary, as a minimum,
40、for the coded picture area to be the same for a given MPEG-2 level so that the MPEG-2 macroblocks within the picture are correctly aligned as defined in SMPTE RP 202. Transcoding techniques now exist to optimally con- vert compressed video bitstreams from one bit rate to another either via digital v
41、ideo baseband or through an I-frame-only stream. These techniques are defined in the MPEG-2 recoding data set, SMPTE 327M. Using recoding information enables the next encoder to follow the same decisions as the previous encoder, thereby minimizing or eliminating concatenation arti- facts. 4.3 Analog
42、 legacy issues SMPTE RP 202 also makes reference to the 512-line and 608-line formats, which were introduced into the MPEG-2 4:2:2 profile to accommodate processing of vertical interval data through the compressed video path. This was done with analog television systems in mind, in advance of substa
43、ntial digital television broadcasting deployment. Particularly in confined data rate environments, mixing data and video paths (composite analog inputs to encoders) has resulted in suboptimal processing of both signals. With increased emphasis on digital broadcast emission, there is new motivation t
44、o pass only video through the MPEG-2 compression process and pass data via a separate data path. In this case, only the 480-line and 576-line formats would be used for standard-def inition televi- sion. Consequently, phasing out the 51 2- and 608-line MPEG-2 formats, which are based only on legacy a
45、nalog requirements, is recommended. Digital high-definition systems do not have the burden of analog heritage of standard-definition systems. HDTV does not therefore have a legacy of analog and digital vertical intervals being carried with compressed video. It is imperative that HDN systems maintain
46、 their clear distinction between video data and ancillary data. 4.4 Standards conversion issues There is frequently a requirement during international electronic program exchange to have both an MPEG- 2 long-GoP link coupled with a frame rate standards conversion process. MPEG-2 relies on smooth mot
47、ion rendition for the accuracy of its predictive modes. Where this motion is modified, as in the case of standards conversion, temporal disturbances will be introduced. These dis- turbances are minimized by the use of motion com- pensated rather than linear standards converters. Ideally, the contrib
48、ution link would be MPEG-2 com- pressed at the source frame rate and standards con- verted at the receiving end, thereby allowing the contribution link to work at its maximum efficiency. This assumes that the contribution link has asuff icient overhead forthe subsequent processing and MPEG-2 recodin
49、g for emission. Although the above situation is preferred, it is acknow- ledged that in the case of point-to-multipoint contribu- tions, the altemative solution of standards conversion before the MPEG-2 contribution link may be chosen for economic reasons. 4.5 Historical process-related metadata As source content is moved through the transmission chain, it is likely to undergo a series of conversions such as coding to MPEGP for a contribution link, standards conversion, conversion from MPEG-2 4:2:2 to 4:2:0 or vice versa. Each new modification is typically performed without referen
