INTERNATIONAL ORGANIZATION FOR STANDARDIZATION
`
`ORGANISATION INTERNATIONALE DE NORMALISATION
`
`ISO/fEC JTC1ISC2IWG8
`
`CODED REPRESENTATION OF PICTURE AND AUDIO INFORMATION
`
`ISO-IEC/JTCIJSC2/WG8
`MPEG 89/128
`July1989
`
`Source
`
`Leonardo Chiariglione
`Tsuneyoshi Hidaka
`James
`Johnston
`Didier Le Gall
`Hans-Georg Musmann
`CoHn Smith
`
`Title
`
`Status
`
`Proposal Package Description
`Draft
`
`Proposal Package Description
`
`1.Purpose and content of this document
`
`The purpose of
`is to summarise in one document
`this proposal package
`reached within the Moving Picture Coding Experts Group MPEG on the method and
`agreement
`time schedule to be followed in order to reach the goal of defining Draft Proposal of standard for
`coded representation of moving picture image and sound information for storage and retrieval on
`digital storage media DSM
`
`description
`
`the
`
`The method adopted relies on the consolidated
`experience in different Standardisation
`Bodies
`including that of the parent Committee WG8
`Call for Pre-registration of intention to submit an
`and 133 has been issued At the deadline specified 39 Companies/Or
`algorithm MPEG88/132
`registered such an intention MPEG89/123 Later 37 were retained and one statement
`ganisations
`of possibility to submit proposal was recorded The different proposals will be evaluated for their
`suitability to the target standard according to the procedure defined below Ultimately
`selection
`will be made which will be the result of collaborative improvements
`
`The time schedule foresees
`reaching the stage of Draft Proposal of standard by 1st September
`1990 This date has been set by JTC1JSC2 at their last plenary meeting
`
`This proposal package description is not self-contained Other documents
`text and are needed in order to fully understand it They are
`
`are referenced in the
`
`MPEG88/001
`MPEG88/132
`MPEG88/
`133
`MPEG89/120
`MPEG89/141
`MPEG89/148
`MPEG89/157
`
`Terms of reference
`
`Call for pre-registration
`Liaison letter
`to related Organisations
`MPEG VLSI Cost estimation Questionnaire
`MPEG image test procedure Rev
`High-Level Description of proposed sound coding algorithms
`kbitls flexible hardware
`Specification for px(cid:243)4
`
`1
`
`HP 1024
`
`

`

`MPEG89/158
`MPEG89/159
`MPEG89/166
`MPEG89/168
`
`Description of Ref Model RM8
`Structure of JPEG Algorithm explanation
`List of agreed audio test sequences
`for subjective testing
`Summary description of the JPEG Baseline Algorithm
`
`livingston
`
`of
`
`consensus
`
`For the purpose of the tests to be carried out in October on Moving Images reference
`Stockholm meeting reports is also recommended
`
`to the
`
`Note
`
`In this text
`
`the foU.owing conventions have been made
`
`before
`
`unit of measure means 1000
`before
`unit of measure means 1024
`before unit of measure means 1000000
`
`The structure of the document
`
`is as follows
`
`Purpose and content of the proposal package description
`list of registered proposers
`Time schedule towards DP stage
`Principles of test methodology
`Implication weights
`
`System requirements
`Performances to be tested
`Definition of test procedures
`Clustering and Preselection
`10 Description of algorithm
`11 Verification of correct generation of test material
`12 Demonstration
`of VLSI implementability
`13 Demonstration
`of basic performance of encoder
`14 Hardware demonstrators
`15 Self scoring policy
`16 Summary of proposal package
`
`2
`
`

`

`List of registered proposers
`
`The following
`is the complete list of those who have registered an intention
`algorithm as per MPEG89/132 and 133
`
`to register an
`
`name
`Company/Organisation
`American Telephone
`and Telegraph
`Bell Communications Research
`BTRL
`CCETF
`Creative Circuits Corp
`COST2llbis
`Deutsche Thomson-Brandt GmbH
`Digital Equipment Corporation
`France TØlØcom CNEF
`Fraunhofer Gesellschaft
`Fujitsu Limited
`Graphics Communications Technologies
`IBM Corporation
`f(cid:252)r Rundfunktechnik
`
`Institut
`
`Intel
`lIT Intermetall
`Japan Victor Company Ltd
`Matsushita Electric Industrial Company
`MiT
`
`Mitsubishi Electric Company
`NEC Corporation
`Nippon Telegraph and Telephone Corporation
`OKI Electric Industrial Company
`Philips Consumer Electronics
`Sony Corporation
`Telenorma /University of Hannover
`Toshiba Corporation
`
`Audio Ref Name
`
`SJayant
`
`Challener
`Y.-F Dehery
`
`Schroeder
`
`Combescure
`
`Gerhaeuser
`
`Unaganii
`
`Stoll
`
`Yamada
`
`Kogure
`
`Nishitani
`
`Hayashi
`
`Lokhoff
`
`Akagiri
`
`Video Ref Name
`Haskeil
`Le Gall
`
`Ligtenberg
`H.Seguin
`Hepper
`Szabo
`
`Texier
`
`Matsuda
`IL Fujiwara
`Gonzales
`
`Simon
`
`Butera
`Hidaka
`
`Kogure
`Lippman
`Murakami
`Omachi
`
`Watanabe
`Kbro
`van der Meer
`Hashimoto
`Kummerow
`
`Sugiyama
`
`Note
`
`Note
`
`Hammer Siemens has stated
`possibility of nialdng
`video satisfying the time scale of para
`
`proposal related to
`
`Two more preregistrations
`for audio were received One was withdrawn and the
`other was rejected on 3rd July for failure to comply with the first requirement
`of para
`
`3
`
`

`

`Time schedule towards DP stage
`
`General
`
`Moving Images
`
`Audio
`
`03/07/89
`
`07/07/89
`
`Weights
`
`finalised
`
`Test procedure finalised
`
`Complexity measure defined
`
`Additional
`
`performances finalised
`
`Proposal
`
`package
`
`finalised
`
`System aspects started
`
`Top level description
`
`delivered
`
`Test procedure finalised
`
`Additional
`
`performances finalised
`
`Test sequences finalised
`
`15/09/89
`
`06/10/89
`
`16/10/89
`
`20/10/89
`
`D-1 tapes delivered to JVC
`
`Full proposal
`
`registration
`
`deadline
`
`Detailed algo description
`
`provided
`
`System aspects defined
`
`Clustering and preselection
`
`of algorithms
`
`Collaborative
`
`algorithm improvements
`
`start
`
`xx/01J90
`
`xxJO2 or 03/90
`
`xx/04/90
`
`xx/05/90
`
`xx/06/90
`
`xx/07190
`
`1/09/90
`Hardware demo of system
`DP ready
`
`Preselection if needed
`
`Fmal selection
`
`Collaborative
`
`improvement
`
`starts
`
`Fully verifiable
`
`package
`
`provided
`
`Test
`
`tapes prepared
`
`Execution
`
`of subjective tests
`
`Final selection
`
`Verification of decoder using hardware
`
`Verif of decoder using hardware
`
`4
`
`

`

`Principles of test methodology
`
`Evaluation of algorithms will be done by means of the formula
`
`Si
`
`Sai Wa
`
`SbiWb
`
`SziWz
`
`where
`
`refers to candidate algorithm no
`
`b..
`
`are measurable parameters
`
`Wa Wb .. Wz
`
`are the relative importances weights
`
`Sai 5bi .. Szi are the corresponding scores for candidate no
`
`Si
`
`is the total score for candidate
`
`Examples of measurable parameters are
`
`quality of video in normal playback
`quality of video in reverse playback
`number of gates/chips
`to implement decoder
`decoding delay
`real time mixing of audio channels
`
`Note
`
`Note
`
`Some implications
`proposal
`
`are mandatoiy Failure to satisfy any of them disqualifies the
`
`For the October
`
`implications
`
`in
`
`tests moving images and the Januaiy preselections audio some
`above will not be used The final selection however will be made
`using all implications
`
`Implications weights
`
`Weights have been derived by
`ballotthig procedure MPEG89/053
`in From ballots of applications
`have taken part
`panies/Organisations
`as indicated below
`averages were calculated
`
`and 108 which 35 Corn
`and implications weights
`
`ImpLi
`
`Impl.2 Impl3
`
`Impl.j
`
`Impl.m
`
`Appl.1 ui
`
`vii
`
`vi
`
`vi
`
`AppL2u2
`
`V2i
`
`v22
`
`VZ3
`
`Appl3 U3
`
`V31
`
`32
`
`V33
`
`Appl.i UI
`
`Vii
`
`Vi2
`
`vu
`
`AppLnu
`
`Vni
`
`vn2
`
`v113
`
`Wi
`
`W2
`
`W3
`
`vij
`
`V2j
`
`V3j
`
`vj
`
`vnj
`
`Wj
`
`Vim
`
`v2m
`
`V3m
`
`Vim
`
`Wm
`
`where
`
`ui
`
`vij
`
`is the relative importance of application
`is the relative importance of implication
`
`for application
`
`5
`
`

`

`Weights
`
`are defined by
`
`Wj jUiVij
`
`Definition of applications/implications
`
`and corresponding weights are given below
`
`5.1 Application weights
`
`Definition of Applications
`
`Electronic Publishing End user
`Integrated and interactive text graphics sound and video
`Electronic
`newspaper
`Electronic Publishing Production
`Digital video editing
`Digital audio editing
`Travel guidance
`Preview of city views
`Flight and driving simulation
`Interactive video in public spaces e.g airports
`Active guidance
`Surrogate travel
`Educational applications
`Electronic book high level of interaction
`Electronic book low level of interaction
`Remote classroom instruction
`Business applications
`Electronic book as l.a for catalog shopping
`Multimedia E-mail as in 9.b
`Games
`Stand alone games
`Multilocation
`games through telecom network
`Entertainment applications
`MTV or other video/audio
`clips on CD
`TV program/movie
`storage and distribution
`Broadband stereo distribution
`Previewing systems both audio and video
`Audio recording
`Database
`applications
`Audio databases
`applications
`Audio
`video either with onsite or offsite databases
`Multimedia memo and bulletin board
`Electronic News Gathering
`Bulletin board
`Computer conferencing with multimedia memos
`using MPEG standard
`Face-to-face
`10 Broadcast sound
`
`teleconferencing
`
`Production
`
`Transmission oftcontribution
`
`quality audio
`
`Transmission for distribution
`
`Digital audio broadcasting
`11 Multimedia Art
`12 Medical Applications
`Education
`
`Diagnostic Imaging
`
`Weight Standard
`
`deviation
`
`3.37
`
`3.04
`
`2.00
`
`2.45
`
`2.67
`
`2.08
`
`2.78
`
`2.78
`
`2.61
`
`3.47
`
`3.26
`
`2.63
`
`3.53
`
`2.51
`
`2.80
`
`2.04
`
`3.12
`
`2.96
`
`2.59
`
`2.55
`
`2.31
`
`2.63
`
`3.00
`
`2.41
`
`2.49
`
`2.43
`
`1.88
`
`2.71
`
`2.78
`
`3.20
`
`3.08
`
`1.69
`
`2.71
`
`2.08
`
`1.73
`
`1.61
`
`1.43
`
`1.46
`
`1.48
`127
`
`1.50
`
`1.54
`
`1.52
`
`1.79
`
`1.64
`
`1.41
`
`1.81
`
`1.43
`
`135
`
`1.43
`
`1.73
`
`1.60
`
`1.73
`
`1.47
`
`1.88
`
`137
`153
`
`1.48
`121
`132
`153
`
`1.65
`
`1.70
`
`131
`
`1.89
`
`1.34
`
`1.64
`
`1.49
`
`6
`
`

`

`5.2 Implication weights
`
`Definition of Implications
`
`Weight Standard deviation
`
`Motion video decoding playback
`Forward video playback
`Reverse video playback
`Fast forward video playback
`Fast reverse video playback
`Wideband audio decoding playback
`Forward audio playback
`Reverse audio playback
`Fast forward audio playback
`Fast reverse audio playback
`Single frame video accessibility
`Random access
`5.Stillmodevideo
`Zoom capability
`Compatibility with computer graphics
`Ability to encode in real thne
`Audio
`Video
`Reduced frame rate video at higher
`resolution
`10 Data capacity for ancillaiy information
`of CCJTF px64 kbitls video
`11 Direct
`12 Direct
`of Studio/Production audio
`13 Direct
`of consumer standard audio
`transcoding
`14 Real
`time mixing of audio channels
`15 Ability to synchronize audio and video signals
`16 Ability to withstand multiple codingldecoding audio
`17 Ability to do insert/cut editing audio
`18 Ability to tolerate audio processing during production
`19 Need of stereo wideband 20 kHz audio
`20 Intermediate audio quality
`21 Toll quality audio
`22 Robustness to bit errors in storage/transmission
`Video
`Audio
`23 Encoder cost and/or complexity
`Video
`
`transcoding
`
`transcoding
`
`audio
`24 Decoder cost/complexity
`Video
`Audio
`25 Ability to handle 20124 bits input audio
`26 Short decoding delay
`Video
`
`Audio
`27 Short encoding delay
`Video
`
`Audio
`28 Direct
`
`transcoding
`
`of JPEG still picture
`
`108
`
`050
`
`091
`
`084
`
`121
`
`034
`
`067
`
`052
`
`078
`
`118
`
`076
`
`049
`
`058
`
`055
`
`058
`
`047
`
`093
`
`026
`
`054
`
`057
`049
`
`109
`
`029
`
`050
`
`031
`
`086
`
`096
`
`054
`
`073
`
`089
`
`042
`
`059
`
`103
`
`117
`
`030
`
`066
`
`072
`
`024
`
`030
`
`24
`
`20
`
`23
`
`24
`
`26
`
`19
`
`21
`
`21
`
`22
`
`26
`
`21
`
`19
`
`19
`
`21
`
`17
`
`23
`
`15
`
`23
`
`23
`
`20
`
`26
`
`17
`
`21
`
`17
`
`23
`
`24
`
`19
`
`21
`
`24
`
`17
`20
`
`23
`
`25
`
`17
`
`20
`
`22
`
`15
`
`18
`
`7
`
`

`

`System requirements
`These requirements must be met by all algorithms submitted
`
`6.1 System aspects
`
`Framing ability for random access/editing/fast
`
`forward etc
`
`6.2 Moving images Not agreed upon on 07/89
`
`Input data format according to
`
`CCIR 601 625/50
`CCIR 601 525/60
`CIF
`
`Output data format according to
`
`CCIR 601625/50
`CCIR 601525/60
`N.B The two outputs have to be provided
`
`irrespective of the format of the input data
`
`Synchronous
`
`and asynchronous coding possible
`
`6.3 Audio
`
`Input sampling rates 3244.148 kHz
`
`Input and output sampling rate should match
`
`Input resolution
`
`16 bit uniform
`
`Bit rates at which the coder should operate
`6496 128 and 192 kbits/sec
`Monophonic
`128 192256 and 384 kbits/sec
`
`Stereo/bilingual..
`
`The algorithms should work at slightly lower rates to allow application specific ancillaly
`data without drastic loss of quality
`
`Performances to be tested
`
`implications whose performance will be subjected
`Listed in the following are all
`to tests All
`scores will be normalised in the range 0-10 with 10 being the highest
`rating Items requiring further
`action are marked with
`
`The order in which implications
`are listed in this paragraph corresponds to that of para
`future it will be organised in system moving images and audio performances
`to be tested
`
`In the
`
`Only the following 13 audio-related implications with weight greater than 55 2a 2c
`13 15 1920 22b 23b 24 and 26b will be considered for testing
`
`8a 10
`
`7.1 Motion video decoding playback
`
`Moving image quality will be tested for the
`basic operating modes Images will have to be
`from the same bitstream The length of the compressed
`decoded
`stream for each sequence will be
`required to be less or equal to the duration in seconds of the sequence times 900 kbits
`
`8
`
`

`

`Normal playback Forward video playback
`Normal playback consists in playing the coded material at the same speed as that of the source
`materiaL At least 704x480 for 525/60 sequences and 704x576
`for 625/50 sequences picture
`elements from the active picture area will be used for coding
`
`Reverie video playback
`Reverse video playback consists in playing the decompressed material at the same speed and in
`reverse order as that of the source materiaL In addition it is required that the maximum allowed
`delay between the time the first byte is decoded from the bitstream and the first picture in reverse
`playback mode be less than is
`
`Failure
`video sequence within the accepted
`to provide
`regardless of the picture quality
`
`delay will warranty
`
`score of
`
`Fastfowazrl video playback and
`Fast reverie video playback
`Fast forward Reverse video playback
`also known as fast search mode
`is defined to be
`playback mode at between
`and 10 times the speed of the original source materiaL The amount
`of data which may be required to be read from the storage media shall not exceed 900 kbits times
`the duration of the fast playback mode The provider will present convincing documentation
`the amount of bits read from the bitstream for that purpose satisfies this constraint
`
`that
`
`7.2 Wideband audio decoding playback
`Foiward audio playback
`Audio quality will be tested by an A-B-X test The input sampling rate used for testing will be
`48 kHz The output bitrates will be for
`Monophonic
`sequences
`Stereophonic or bilingual sequences
`
`96 and 128 kbit/s
`192 and 256 kbit/s
`
`The originally proposed quality score for algorithm
`
`sqi
`
`101-Ndi/Ni
`
`where Ni is the number of times subjects heard coder
`subjects heard differences in coded/original
`pairs from coder
`the A-B test
`to the A-B-X test
`
`and Ndi is the number of times
`has to be adapted to the change from
`
`of Ndi as the number of correct assignments of
`Interpretation
`to either
`or
`appears not
`to be appropriate Therefore the members of the audio group are asked to submit proposals for
`quality score sqi until September 301989 These proposals will be distributed
`more appropriate
`by fax before the next meeting
`
`Reverie audio playback
`This implication will not be tested
`
`Fastfoiward audio playback
`forward audio playback is defined to mean that
`Fast
`of evely
`coded audio frames will be
`dropped from the bit stream and the decoder will have to do something reasonable with the
`bitstream This may include
`providing speeded up sound for fast
`forward.There are two initial
`drop access units to get 4x speed and
`proposals
`drop access units corresponding to dropped
`video access units
`
`forward audio playback will be tested by applying
`the CCIR 562 quality scale based on
`Fast
`computer simulationsThe resulting scoring scale
`to will be mapped linearly to the range
`to
`10
`
`cL Fast reverie audio playback
`
`9
`
`

`

`This implication will not be tested
`
`3.Single frame video accessibility
`
`Video
`While the two implications
`for the purpose of testing performance they are the
`are different
`same No subjective testing will be carried out for those implications
`since as is quality was
`assumed in the implication ballot An objective measure for single frame accessibility
`is thus derived
`as follows
`
`The measure of single frame random access
`is defined as the maximum time elapsed from start
`to finish of reading from the DSM any frame following
`request it is clearly related to the maximum
`in bits or in seconds at
`the rate of 900 kbit/s
`between breaks in interframe dependency
`
`distance
`
`random access
`time of Os will warrant
`random access
`time
`Any random access
`
`2s will warrant
`in the range 02 will
`
`full score 10
`zero score
`
`lead to
`
`score
`
`10-5x
`
`in seconds
`
`The proposer will provide details of breaks in the interframe dependency
`for all three sequences
`and calculate the worst case delay or provide an equally convincing alternative
`
`Random access
`CL Video
`See 73 above
`
`b.Audio
`Neeth clarification
`
`7.5 Still mode video
`This will be measured as still image quality obtained 0.3
`after the moving image is stopped
`This supposes that an additional 270 kbits is available for higher resolution information beyond the
`bitstream described in 7.1 No limitation on the kind of algorithm enabling the improvement from
`the lower resolution moving image to the higher resolution
`still are imposed however
`understood
`that any additional
`decoding device will affect each proposal score on item 24.a
`decoder cost/complexity
`
`it is
`
`7.6 Zoom capability
`This capability will not be tested since it addresses
`
`capability very similar to 7.5 above
`
`7.7 Compatibility with computer graphics
`Weeds clariflcan
`
`7.8 Ability to encode in real time
`CLAUthO
`Neeth clarification
`
`Video
`This capability as well as measure of the encoder complexity will not be tested in October 1989.
`
`7.9 Reduced frame rate video at higher resolution
`sequence temporally sampled @5/6 Hz 1/5 the
`This implication is measured
`as the quality of
`frame rate of CCIR 601
`In view of the expected applications of such mode the temporal correlation between adjacent
`frames should be high This capability will not be tested in October 19989 in view of the difficulty
`of providing relevant coding material
`
`10
`
`

`

`7.10 Data capacity
`CL Video
`
`for ancillary information
`
`b.Audio
`The maximum expected bitrate for audio related ancillary data is
`transmission bitstream
`
`bit per every 64 bits of the
`
`All subjective tests will be performed with zero ancillary data
`
`transcoding of CCITT px64 kbit/s video
`7.11 Direct
`binary decision direct transcoding means that the MPEG decoder
`This will be measured as
`able to decode exactly CCITT bitstream Fig 7.a illustrates the procedure of transcoding
`since
`this is widely viewed as
`hardware feature the transcoding
`circuitry if any is needed will affect
`the decoder cost/complexity
`as per item 24.a
`
`is
`
`Annex
`
`gives the latest CC1IT documents
`implication
`
`that can be used for the purpose of assessing this
`
`7.12 Direct
`transcoding of Studio/Production Audio
`This implication wifi not be tested
`
`7.13 Transcoding to AES/EBU consumer standards and CCITT video conferencing standards
`44.1 kHz sampling rate and 48 kHz sampling rate The
`Consumer
`standard audio includes
`problem of transcoding will not be part of the proposal package
`
`7.14 Real time mixing of audio channels
`This implication will not be tested
`
`7.15 Synchronization of audio and video
`This is
`system issue which cannot be tested without system level
`Needs clanfication
`
`information
`
`7.16 Ability to withstand multiple coding/decoding audio
`This implication will not be tested
`
`7.l7Abffity to do insert/cut editing audio
`This implication will not be tested
`
`7.18 Ability to tolerate audio processing during production
`This implication will not be tested
`
`7.19 High quality stereo
`Needs clarification
`
`7.20 Intermediate quality audio
`The same test procedure as for forward audio playback 7.2a will be applied at an output bitrate
`of 64 kbit/s
`
`7.21 Toll quality audio
`This implication will not be tested
`
`7.22 Robustness to bit errors in storago/transmission
`Video
`Robustness to bit errors in storage/transmission
`however will have to check this implication
`
`will not be tested in October The final
`
`tests
`
`11
`
`

`

`b.Audio
`
`According to
`
`proposal of the Rennes meeting the coders will be rated with random bit errors
`inserted in the coded bitstream Suppose we is the weight
`from the ballots
`for error tolerance
`MPEG89/077 and seri is the quality score for algorithm under
`random errors
`
`The overall score for algorithm
`
`Si
`
`.. weseri
`
`will be computed as in MPEG Doc 89/041
`
`lithe scores for two algorithms are statistically the same then the quality score under byte errors
`sebi will be used to compute the overall score
`
`Si
`
`.. wesebi
`
`and used to rank the algorithms if these scores are statistically different
`
`23 Encoder cost and/or complexity
`Video
`No score will be attached to this feature in October It should be kept in mind that this parameter
`makes sense only for an encoder that has some real time encoding capability since in theory trivial
`encoder 4-bituprocessor could achieve
`the desired result in an extremely long time
`
`Audio
`It has been agreed to adopt
`
`the proposals
`
`that will be provided by the VLSI group.
`
`7.24 Decoder
`cost/complexity
`See report of the VLSI implementation Ad-hoc Group
`
`7.25 Ability to handle 20/24 bit input audio
`This implication will not be tested
`
`7.26 Short decoding delay
`Video
`In the framework of decoder working from stored bitstream careful analysis shows that the
`decoding delay is bounded above by the random access
`described in 7.3 7.4 Since the
`time
`weight associated with 7.3 and 7.4 is particularly high no conclusive
`evidence was given that
`feature needs to be tested independently
`
`this
`
`Note
`
`Note
`
`An algorithm that does not voluntarily place breaks in the interframe
`dependency
`could score highly on 7.26a and poorly on 7.3
`7.4 There was little support
`to reward such an algorithm with additional points
`however
`The decoding delay for video is however an important system parameter since
`video and audio are expected to be synchronised
`
`Audio
`The following scoring has been agreed
`s10
`
`101-tcj-201230
`
`s0
`
`where t1
`
`delay in ms
`
`for0ti20
`
`for20
`
`tj 250
`
`fortd250
`
`12
`
`

`

`7.27 Short encoding delay
`Jdeo
`Neeth clarification
`
`Audio
`This implication will not be tested
`
`transcoding of JPEG still pictures
`7.28 Direct
`binary decision direct transcoding means that the MPEG decoder
`This will be measured as
`JPEG bitstream Fig 7.b illustrates the procedure of transcoding
`able to decode exactly
`this is widely viewed
`hardware feature the transcoding
`as
`circuitry if any is needed will affect
`as per item 24.a
`the decoder cost/complexity
`
`since
`
`is
`
`Annex
`
`agreement
`
`by the JPEG Chairman following
`document
`II provides
`reached during the last Stockholm WG8 Plenary
`
`reference
`
`supplied
`
`the
`
`Definition of test procedures
`
`8.1 Moving images
`
`Test Material
`
`Table Tennis
`
`duration
`
`10
`
`tested in
`
`Flower Garden
`
`duration
`
`tested in
`
`Diva
`
`duration
`
`tested in
`
`Flower Garden HRS duration
`
`is
`
`tested in
`
`normal playback
`reverse video playback
`fast forward/fast
`reverse
`normal playback
`reverse video playback
`normal playback
`reverse video playback
`high resolution still mode
`
`Reference Material
`
`CCIR 601
`
`720x240
`
`use for
`
`horizontal
`LPF 1/41/21/4
`
`1st field
`normal playback
`reverse video playback
`fast forward/fast
`reverse frame sampling corresponding to each proposal
`
`vertical
`Lin Interp
`
`720x480 1st
`2nd
`
`odd
`
`even
`
`maximum resolution
`CCJR 601
`use for still mode video
`
`Simulation reufr Record foll(cid:224) wing items to 1-i tape
`
`Normal playback
`Reverse video playback
`Fast
`forward/fast
`reverse
`
`Still mode video
`
`lOs Table Tennis
`
`lOs Table Tennis
`
`combination of
`
`Ss Flower Garden
`5s Flower Garden
`Diva
`is of fast forward and is of fast reverse
`
`Diva
`
`times
`repeat
`is Flower Garden more than lOs of still picture
`
`lVstandard for October tests
`
`Simulation results shall be recorded in 525/60 or 625/50 or both
`At the October meeting two algorithms will be selected
`from proposals using the 625/50
`from proposals
`format and four
`using the 525/60 format
`
`13
`
`

`

`Countries using the 625/50 standard may send their processing
`tapes COST 211 format JVC will convert
`them to D-1 format
`
`results in 525/60 on magnetis
`
`Method of assessment MPEG89/141
`
`continuous quality method will be used
`The double-stimulus
`The number of assessors will be 255 stations each having
`assessors
`Each proposer will be entitled to nominate one assessor
`In case of proposals originated by
`maximum of three assessors will be nominates If the total
`group of companies/organisations
`number will exceed 25 drawing will be made If the total number will be lower than 25 MPEG
`members having attended at least
`two of the Uvingston Rennes Stockholm and Tokyo meetings
`will be admitted
`
`Schedule
`
`Deliveiy of test material
`JVC will send test material on computer
`525/60
`ATT
`Japanese proposers
`by July 15
`ATT will distribute do to
`
`tapes to
`
`Beilcore
`
`Creative Circuits Corp
`DEC
`IBM
`
`625/50
`
`Intel
`Mif
`DTB
`by July 20
`LER will send test material on computer
`BTRL
`DTB
`rrr
`
`tapes to
`
`by July 15
`BTRL will distribute do to
`PRL
`COST211
`CNET
`by July 20
`DTB will distribute do to
`SELT
`LEP
`Univ Hannover
`by July 20
`iT will distribute do to
`SIEMENS
`TELENORMA
`ATT
`BTRL will distribute do to
`625/50 D-1 DTB will send test material on D-1 tape to
`JVC
`
`by September
`
`15
`
`Simulation results on D-1 tape sent to JVC by September 15 results on computer tapes only
`via express courier 48 hour deliveiy
`for European proposers
`by September
`
`14
`
`

`

`Assessment schedule
`
`for October testc
`
`October
`
`October
`
`17 rue
`18 Wed
`19 fhu
`20 Fri
`
`525/4
`
`625/50 SFD
`
`Bitrates to be used for simulations
`
`The normalised bitrates of 1150 and 900 kbit/s will be used However
`the latter will be used
`
`for the October
`
`tests only
`
`8.2 Audio
`
`Test Material
`
`Twenty test sequences have been selected during the Stockholm meeting The list of the selected
`test sequences will be provided in separate MPEG document
`
`Reference levels will be set at 90 db and 110 db SPL Each test sequence should be anchored
`to these levels The test material will be sampled at 48kHz and provided on DAT
`with respect
`
`In the case that algorithms will be tested which exploit the stereo channel
`additional
`test sequences and proposals
`test procedure are needed
`for
`
`interdependency
`
`some
`
`Test Methodology
`
`minimum of 30 subjects will be used The subjects will be experts listeners The test material
`will be made available to the subjects for training before the final test However
`this material will
`be in different order from the test tape Algorithm proposers will have the right to nominate
`subjects each Other committee members can nominate
`subject per company AES/EBU will
`call for other subjects
`
`During the Stockholm meeting it has been agreed to change the procedure for subjective testing
`of forward audio playback quality from the double blind forced choice A/B test
`to an A-B-X test
`The members of the audio group are asked to prepare proposals
`for an appropriate
`and statistical evaluation of the A-B-X test by the October meeting
`
`test procedure
`
`Testing should be done at neutral places Proposals for locations of subjective tests or volunteers
`the October meeting
`are requested until
`
`Both headphone and loudspeaker
`tests will be done For headphone tests Stax Lambda Pro
`headphones with diffuse field equalizers will be used Loudspeakers will be selected at the place
`of the tests For loudspeaker
`tests the number of subjects listening at one time will be limited to
`to ensure that all subjects hear the same thing
`
`The reference
`dBand 110dB
`
`level calibration will be done using 113 octave pink noise centered at
`
`kHz at 90
`
`Preselection
`
`9.1 Moving images
`
`9.1.1 Scoring method and weights for items to be tested in October
`
`15
`
`

`

`Forward normal playback
`Reverse normal playback
`forward and reverse
`Fast
`Still images with high resolution
`Random access figure
`of CCITt
`Direct
`of JPEG
`Direct
`transcoding
`VLSI implementabifityof decoder
`
`transcoding
`
`9.1.2 Clustering
`
`scoring method
`quality scale
`quality scale
`not scored
`
`quality scale
`maximum time
`or 10
`
`binaiy score
`
`binaiy score
`implementation cost figure
`
`or 10
`
`weight
`108
`
`50
`
`76
`
`118
`
`26
`to be sent by 15/08/89
`103
`
`Proposals will be clustered according to the following figure of merit
`
`jSjWi
`Score Quality Nornnl Playback
`Score Quality Reverse Playback
`Score Quality Hi-Res Still
`Figure of Merit Rdom Access
`Score Transcoding ccrrr
`Score Transcoding JPEG
`Score VLSI implementability
`
`108
`
`50
`
`76
`
`118
`
`26
`
`103
`
`9.2 Audio
`
`9.21 Pnselection
`
`ndes
`
`In the audio group it was decided to cluster the proposed audio coding algorithms into the fol
`lowing groups
`
`Transform coding with overlapping blocks
`Transform coding with non-overlapping blocks
`Subband coding with less than
`subbands
`Subband coding with more than
`
`subbands
`
`In order to reduce the amount of testing in the final selection procedure the following recom
`mendations for the audio group have been stated
`
`The members are encouraged
`to join one of the
`groups
`Members who are not willing to join one of the
`groups have to indicate their intention
`to proceed with an individual
`algorithm by the October meeting
`If there are individual
`can be installed during the October
`algorithms
`pre-selection
`meeting by majority vote of the proposers
`The preselection
`is based on the following performances
`
`quality
`decoder complexity
`decoding delay
`algorithms will be considered which are better
`After preselection only those individual
`to the worst collaborating
`or equal
`group of algorithm
`The same procedure applies if there is
`smaller number of groups
`
`After preselection only those individual algorithms will be considered which are better or
`equal to the algorithm with the lowest performance of
`
`collaborating group
`
`16
`
`

`

`9.2.2 Peifonnances to be tested forpresekction
`
`For preselection simplified tests will be performed and only three implications will be con
`sidered
`
`Forward audio playback
`The same test as described
`channel
`
`independednt
`
`in 7.2.a will be applied
`
`at only one output bitrate of 128 kbit/s per
`
`Decoder cost/complexity
`As long as there is no specific recommednation available from the VLSI group the following
`simplified evaluation of the VLSI questionnaire
`will be applied
`
`The silicon area
`
`required for the decoder
`
`is to be estimated according to the rules of the
`
`questionnaire
`To the decoder with the smallest area Amin the score 10 is assigned and to that of the maxi
`mum area Amax the score will be assigned All other decoders will be scored linearly according
`to this range
`
`Decoding delay
`The same test as described
`
`in 7.26.b will be applied
`
`10 Description of the algorithm
`
`algorithms will be described by means of block diagrams These will have to describe
`Proposed
`also how additional performances can be obtained
`
`For the audio group the detailed description of the algorithm as is to be provided by October
`16 1989 is understood as follows
`
`Detailed description means that enough detail
`is to be provided in the description to enable
`other parties to reproduce the results Questions may be directed to the proposers until December
`311989 Proposals may be enhanced
`and similar proposals may be merged between October 89-
`March 90 proposers are asked to indicate their intentions with regard to intellectual properties
`
`11 Verification of correct generation of test material
`
`11.1 Moving Images
`
`Proposers will have to supply
`
`Full description of decoder software by means of flow charts
`
`Executable
`
`code of decoder
`
`Files of bit streams generated by encoder
`
`Files of decoded images
`
`Those proposals which will survive the October
`tests will be verified In case the executable
`code
`can be run on
`the laboratoiy of another MPEG member
`computer which is also available at
`the
`test will be made running the executable
`code of the decoder using the flies of coded data as input
`and checking that the images so generated are the same as those produced by the proposer
`
`17
`
`

`

`In case no other similarcomputer can be found MPEGwill identify reviewers who will undertake
`the writing of the decoder program and perform the same tests as above
`
`112 Audio
`
`Proposeis will have to supply
`
`Full description of decoder software by means of flow charts
`
`Executable
`
`code of decoder or hardware
`
`Files of bit streams generated by encoder
`
`Files of decoded audio tests sequences
`
`12 Demonstration of VLSI implementability
`
`tests moving image decoder VLSI implementability may be
`For the purpose of the October
`submitted Proposers will have to provide
`block diagram of decoder
`system architecture
`hardware reference model
`
`The scope of this block diagram should be as follows
`
`Assume an asynchronous data input stream coming from the Digital Storage
`Audio Decoding
`Media All functional
`input buffering should be shown up to but
`blocks
`including any necessary
`not including the Digital to Analog converter It is assumed that all proposers will use similar
`to
`and circuits
`relating to this so these do not have to be shown in definition of algorithm
`converters
`
`Video Decoding
`Assume an asynchronous data input stream coming from the Digital Stor