&%PAGE& - &%page& - Annex 1 to Doc. AVC-106R Annex 1 to Doc. AVC-106R Documents for the third meeting of the Experts Group 14-23 August 1991, Santa Clara Normal Documents AVC-65R REPORT OF THE SECOND MEETING OF THE EXPERTS GROUP FOR ATM VIDEO CODING IN PARIS (CHAIRMAN) Achievements and action points obtained at the second meeting are recorded to facilitate our discussion at this meeting. AVC-66 REPORT OF THE 14TH MPEG MEETING IN PARIS (MPEG) This document contains the recommendations of the Paris meeting and detailed meeting reports. AVC-67 LIAISON STATEMENTS AND MEETING NOTICE (SPECIAL RAPPORTEUR OF TG CMTT/2) Functional requirements for the secondary distribution of digital CTV, EDTV and HDTV signals to consumers are given on the following aspects; - real-time coding/decoding - minimum coding/decoding delay - automatic optimization to the picture content, network performance, etc. - picture formats covering CTV interlaced, EDTV/HDTV interlaced and progressive - quality objectives - no impairments in principle - graceful degradation for unusual significant level of errors - quality of normal real time delivery be not compromised for features - compatibilities among CTV, EDTV and HDTV - worthwhile commonality with other standards AVC-68 LIAISON STATEMENTS AND IVS BASELINE DOCUMENT (SGXVIII) Outcome of the June meeting which is relevant to our work is contained; status of discussion on CLP issues, status of agreements on UPC issues, and updated IVS Baseline Document. AVC-69 CLR AND BER PROTECTION FOR ENHANCED END-TO-END USERS QOS (RTT BELGIUM) An example of AAL Type 2 implementation is presented, which protects against bit errors and cell losses to obtain at least 2 hours errors free transmission for a bit rate range of 2 to 140 Mbit/s for BER=1E-6, CLR=1E-6. Bit error protection is with forward single error correction, while cell loss protection is with 4 bit Sequence Number and one parity cell added to the 31 service cells. Three critical cases are also analyzed which require slightly smaller CLR. AVC-70 UPDATED DRAFT PROPOSAL PACKAGE DESCRIPTION FOR MPEG PHASE 2 (CHAIR OF MPEG REQUIREMENTS SUB-GROUP) This is an update of the Paris meeting outcome based on the comments received from the collaborators. AVC-71 THE IMPACT OF THE NETWORK ON VIDEO CODING (AUSTRALIA) It is claimed that optimum performance of ATM video coding can not be obtained if source coding and channel coding are considered separately, and that the following ATM network characteristics should be taken into account for development of the ATM video coding; - cell loss resilience - effective use of cell loss priority - variable bit rate coding - flexible bandwidth allocation AVC-72 REQUIREMENTS FOR ATM VIDEO CODECS (AUSTRALIA) The following additions to the list of requirements on the ATM video codecs are proposed; 1) functional requirements - audio/video relative delay (portion assigned to codec) - support of continuous presence multipoint 2) compatibility / interworking requirements - capability of handling multiple coder input/output formats - backward/forward compatibilities with H.261 - interworking among systems with different spatial and temporal resolutions 3) network related requirements - rate control for VBR - control for two priority channels - error performance assessment under cell loss It is concluded that the full list of requirements be input to the IVS Baseline Document. AVC-73 ARCHITECTURE FOR VIDEO SERVICE PROVISION (AUSTRALIA) Flexible layering scheme is supported for the ATM video coding architecture, where signals are coded in multiple layers but lower layers may be switched off as necessary, thus providing the compatibility with existing coding schemes in the short term while allowing the coding scheme optimized for the B-ISDN. A three layer system with pyramidal decomposition of the input signals is given as example implementation. The importance of reaching a decision on the coder input formats defining layer boundaries is stressed by listing issues which impact on the decision; spatial resolution and frame rates, interlace or progressive scanning and pel aspect ratio. AVC-74 THE EFFICIENCY OF A LAYERED PYRAMID CODER (AUSTRALIA) Coding results are provided for a layered Laplacian Pyramid scheme, where the input CIF signal is decomposed into a QCIF base layer and an enhancement layer. Coding efficiency is compared between the single layer scheme and the Laplacian Pyramid scheme in terms of rate-distortion, where rate is counted as entropy of non-zero coefficients. It is concluded that the Laplacian Pyramid scheme is potentially as efficient as the single layer coder. AVC-75 SOURCE/NETWORK MODEL (AUSTRALIA) To improve accuracy of the network loading model, use of cell loss ratio CLR given in Eq.(2) instead of saturation probability Psat is proposed. Number of channels and statistical multiplex gain (SMG) are exemplified assuming a network model where all the cells exceeding network capacity are lost and a source model of Gaussian distribution, concluding that the Group's first network loading model provides conservative estimation of SMG. It is also stressed that understanding of the impact of the network characteristics on the coder will require a model of the cell loss burst behavior. AVC-76 VIRTUAL PATH SUPPORT OF MULTIMEDIA MULTIPLEXED (AUSTRALIA) The Virtual Path approach for ensuring cross media synchronization is concluded as not appropriate because of its semipermanent nature and mixed QOS requirements corresponding to respective media. It is proposed to make a request for the network to provide necessary signalling and control for bounded cross media delay by ensuring all media components follow a common physical path. This should be liaised to SGXVIII for action. AVC-77 THE SAR SUBLAYER OF AAL2 FOR VIDEO SERVICE SUPPORT (AUSTRALIA) Functionalities of SN, IT, LI and CRC in the current I.363 are reviewed from the VBR video service provision point of view, indicating that they may not be matched to the needs of video services, or even not required at all and suggesting that AAL Type 2 with minimum functionality may offer the flexibility to accommodate a wide and diverse range of video services. It is proposed that this matter be a priority issue of the Experts Group and that consideration should be given to a single SAR sublayer for both VBR and CBR with a liaison to SGXVIII advising our thought. AVC-78 REMOTE VIDEO SURVEILLANCE AS A B-ISDN SERVICE (AUSTRALIA) It is proposed that remote surveillance be added to the list of relevant applications being considered by the Experts Group (Table 1 of Status Report), as its characteristics are not entirely covered by other services. It is noted that the remote surveillance has such distinct characteristics as importance of ensuring a high coding quality for each frame for archiving and later retrieval, rapid removal of transmission error artifacts, wide variety of frame rates, frequent use of on-line panning, tilting, zooming, etc. AVC-79 LOSS OF CODING EFFICIENCY BY USING SCIF (JAPAN) Simulation results are reported for comparison of rate-distortion between coding the original 525/60 interlaced signals and coding the converted signals into progressive scan SCIF as in AVC-29. The loss of coding efficiency is found to be 25-30% for INTER and 35-30% for INTRA of "Flower GArden". It is concluded that further study is required, including the use of suitable VLC and/or better format conversion filters. AVC-80 CONVERSION SIMULATION BETWEEN 525 LINE INTERLACED FORMAT AND SCIF (JAPAN) Simulation results are reported on conversion from 525 line interlaced picture to 576 line progressive scan picture and back to the original interlaced picture. Used filters have 16 taps for interlace/progressive scan conversion and 31 taps for 480/576 line conversion. It is concluded that SNR and subjective investigation show satisfactory re-converted pictures but that further study is required on coding efficiency and additional hardware for adopting the SCIF approach. AVC-81 PROGRESSIVE SCANNING CAMERA (JAPAN) Present status for the development of progressive scan cameras in Japan is described as being still in an infant stage. AVC-82 MERITS AND DEMERITS OF THE PROGRESSIVE AND INTERLACED FORMATS (JAPAN) Two possible scanning systems for SCIF, interlace and progressive, are compared on the following items; - ease of layered coding - matching with current camera/display devices - coding efficiency - field parity identification AVC-83 COMPARISON OF PICTURE QUALITY AND ENCODING EFFICIENCY BETWEEN SCIF - SUPER CIF - AND INTERLACED FORMAT USING MPEG ENCODING (JAPAN) Simulation results are shown which compare the coding efficiency between the progressive (SCIF) and interlaced (ILF) formats by applying MPEG encoding separately to each original field. It is reported that though the SCF requires considerably more bits for a common step size (thus obtaining higher SNR) than ILF, both provide almost the same SNR for a common bit rate 4 mbit/s. A possible interpretation is also given. AVC-84 ADDITIONAL SUBMISSION MATERIALS FOR 'KURIHAMA TEST' (JAPAN) Considering that cell loss and its effect on video coding in the B-ISDN ATM environments, it is proposed that the following additional materials be submitted for the Kurihama test; - what will happen if cell loss occurs during transmission, description of cell loss protection, if any - additional delay due to the above cell loss protection - presentation of decoded pictures when cell loss occurred The first two should be mandatory, and the third be desirable. AVC-85 CONSIDERATION ON ACCEPTABLE PROCESSING DELAY IN THE VIDEO CODEC (JAPAN) Starting from the G.114 specifying that one way propagation time should be less than 400ms for "acceptable" audio quality, and considering that the acceptable delay for videophone is not different from telephone, acceptable codec delay is derived for three cases of transmission: optical, satellite and their combination with assumption of 50ms delay due to the ATM processing in the network portion. AVC-86 REQUIREMENTS FOR THE VIDEO CODING STANDARD H.26X (JAPAN) Expected terminals functions in the B-ISDN are first discussed from service integration, namely H.32X intercommunication with another H.32X, MCU, secondary distribution center and network database, as well as multipoint communications point of view. Then the following requirements are derived from them necessary functions to effectively utilize the B-ISDN; 1) H.26X video coding should incorporate H.261 for intercommunication. 2) It is sufficient in the terminal that only the decoder can cope with required service characteristics except for the case of conversational services. 3) Decoders may have full set of capabilities to fulfill the entire functionalities corresponding to respective services, but coders may implement limited set of functionalities specific to the application. 4) Embedded bit stream structure is a desirable coding architecture. 5) Realization of "Flexible Layering" can be a solution for these requirements. 6) Several received video streams may desirably be processed with simultaneous decoding (time-sliced decoding) at the decoder for continuous presence. AVC-87 CONSIDERATION ON MULTI-POINT COMMUNICATION IN B-ISDN (JAPAN) Three configurations for the continuous presence multipoint system are considered, out of which the following two cases need particular attention for the video coding; - A decoder receives multiple bit streams multiplexed from multiple sources and reconstruct respective pictures and provide them to separate displays (time sliced decoding). - A decoder receives a bit stream which contain a picture composed by an MCU from multiple sources. Coded bit streams should be edited without decoding and recoding at the MCU. AVC-88 THE EVALUATION OF SUBBAND CODING EFFICIENCY (JAPAN) Coding results are presented in rate-distortion curves and processed pictures which compare a two layered subband coding and a non-layered coding. The following observations are given; - The quantizer step size for N layered subband coding should be SQRT(N) times smaller than the step size of non-layered coding to provide the same SN. - Adjustment of MC size, MC precision and number of interpolation filter taps improves the subband coding efficiency. - The single layered coding performs better in rate-distortion characteristics, but the reconstructed pictures look almost the same. AVC-89 CONSIDERATIONS ON THE WINDOW SIZE OF TRAFFIC DESCRIPTOR (JAPAN) VBR coding simulation results are given when a particular long term average control scheme is applied to the coder. It is reported that several seconds are required in this coding control scheme for the window size of the traffic descriptor in the network to absorb the changes of of video sequences due to zooming of two seconds. AVC-90 CONSIDERATION ON DELAY WITH VBR CODING (JAPAN) Coding/decoding delay is analyzed for a VBR system with some assumptions of simplification. The size of the encoder output buffer is given as Sbuf(bits)=Ave(bit/T)*Swin(T) where Ave is average bit rate per frame period T and Swin is window size of the network UPC measured in frame time period T. Average bit rate is policed with moving window of size Swin here. Maximum delay is given as L=(Ave/Peak)*Swin where Peak is peak bit rate per frame period T. Some comparison with the discussion in AVC-56 is also given. AVC-91 MULTIMEDIA MULTIPLEX METHOD (JAPAN) Multimedia multiplex methods are discussed in the light of interworking between an H.32X terminal connected to B-ISDN and an H.320 terminal connected to N-ISDN. Some problems of using the VCI method for such interworking are first listed; low bit rate for each medium in H.320, speed of configuration change, provision of multimedia multiplex conversion in the gateway. Then user multiplex is suggested for such purpose, but pointing out the need of study on how to achieve VCI based multimedia multiplex with the interworking feature. Notion of "switchable" and "embedded" is illustrated for multimedia multiplex. AVC-92 VIRTUAL LAYER APPROACH FOR MULTIMEDIA MULTIPLEX FUNCTION (JAPAN) The virtual MUX layer approach is further discussed where the MUX layer is placed between codecs and the network (AAL) for logical multimedia multiplexing functions as presented in AVC-53. Necessary functions of this MUX layer are listed; such as packetization in variable length of each media information, recovery of end-to-end synchronization of each media information, handling of AAL-SDU for lost and errored cell, etc. Mapping of these MUX layer functions to AAL functions are also given. AVC-93 REQUIREMENTS TO THE FUNCTION OF AAL TYPE 1 (JAPAN) Requirements to the CS layer functions of AAL Type 1 are proposed for liaison to SGXVIII in the light of video transmission and its impact to AAL Type 2. - error correction: interleaving should be optional, cell loss notification is indispensable for the video decoder. - 8 kHz timing is necessary for interworking with existing terminals - clock: inquiry should be made whether submultiples of 155.52 MHz are available, recovery of video signal clock may be required. AVC-94 COMPATIBLE CODING OF CCIR 601 IMAGES: PREDICT THE PREDICTION ERROR (PTT RESEARCH - THE NETHERLANDS) Modifications to the multi purpose coding scheme, which has upward and downward compatibility with H.261 or MPEG1 and was presented in AVC-33, are provided with D1 tape demonstrations for both of full CCIR 601 resolutions and SIF resolution results. A second DPCM loop operating on CCIR 601 resolution is added so that two prediction errors are computed: one based on a SIF prediction and one based on the full resolution, the SIF quantized prediction error signal being used for prediction for the full resolution prediction error signal. By switching off the first loop, this system can operate in a non-compatible mode as well. AVC-95 ESTABLISHMENT OF NEW SC29 (CONVENOR OF ISO/IEC JTC1/SC2/WG8) It is announced that SC29 has been established which covers audio, visual and multimedia coding fields. AVC-96 STATUS REPORT ON ATM VIDEO CODING STANDARDIZATION (EXPERTS GROUP) Study results as well as further study items are summarized for common understanding among members of the the group and publicity of the group's activities. This is a version updated according to the Paris meeting outcome. AVC-97 COMPARISON OF NETWORK LOADING MODELS FOR TWO LAYER VARIABLE BIT RATE CODING (UK) It is concluded that the Large Deviation model adopted in the Group's first network loading model is more appropriate than other ones in that it is a safe conservative model independent of the source modeling and that the practical network resource management is more correctly reflected, by giving an equation for the cell loss ratio CLR instead of the probability of saturation Psat. A comparison of number of accommodated channels is given between estimates based on CLR and Psat, indicating that the use of Psat is practically sufficient when about 1000 channels are multiplexed. AVC-98 REQUIREMENTS ANALYSIS FOR VIDEO SERVICES (RTT BELGIUM) Impacts of video service requirements on ATM network parameters and required error handling are discussed for possible upgrading of the IVS Baseline Document. Based on the necessary error free time for each service, bit error handling is first investigated with the ATM payload scrambling taken into account. It is concluded that some bit error protection is mandatory for video services even in case of BER=1E-8, and that the payload scrambling weaken the error correction capability of existing codes such as BCH(511,493) of H.261. Then cell loss handling is investigated, concluding that at least its detection is essential, and that a single cell loss correction is required for the most demanding TV distribution services even for CLR=1E-8. Comments are also given to possible interaction between bit error correction and cell loss correction, bursty cell loss correction and cell loss concealment. AVC-99 VBR VIDEO STATISTICS (RTT BELGIUM) The number of 112 bit subpackets per frame are recorded for the four different type of video services; video telephony, videoconferencing, normal quality broadcast video, high quality broadcast video. An adaptive DPCM codec is used for this measurement which adaptively selects intrafield prediction, interframe prediction and conditional replenishment. This paper is accompanied by a floppy disk containing the recorded data. AVC-100 LAYERED CODING FOR COMPATIBILITY (UK) Coded results are presented for a "pel split" layered coding scheme which uses the MPEG-1 coded pictures as a prediction for the MPEG-2 pictures and has three possible prediction modes for each field; MPEG-1, previous field and previous frame. MPEG-1 bit stream is embedded in the composite bit stream. By switching off the MPEG-1 prediction mode, this scheme becomes a MPEG-2 single layered scheme. AVC-101 COMMENTS ON THE IVS BASELINE DOCUMENT AVC-25 (BELGIUM, FRANCE, FRG, NORWAY, THE NETHERLANDS, SWEDEN, UK) Comments on the following items are proposed for upgrading the IVS Baseline Document; - work plan: consideration of video service requirements such as in AVC- 98 when defining service dependent CS - service bit rates: concern with CLP bit change by the network - QOS provided by the ATM layer: clarification needed - general: stress on multimedia system aspects required AVC-102 COMMENTS ON PICTURE FORMAT - ANNEX 3 TO DOC. AVC-65R (BELGIUM, FRANCE, FRG, NORWAY, THE NETHERLANDS, SWEDEN, UK) It is proposed that one additional format is required beside QCIF and CIF for communication services, whose parameter values are as in AVC-29, awaiting firm supporting evidence concerning quality degradation loss of coding efficiency due to scan conversion by the beginning of 1992. It is also proposed that a future standard should leave the door open for use of other formats. AVC-103 SIMULATION RESULTS ON A PEL SPLIT COMPATIBLE ALGORITHM (CNET - FRANCE TELECOM) Coded results are presented for a "pel split" scheme as summarized above (AVC-100). Numerical data and processed pictures are given for MPEG-1 coding, MPEG-2 compatible coding and MPEG-2 non compatible coding. AVC-104 SUMMARY DESCRIPTION OF VADIS/COST ALGORITHM GROUP 2 FORWARD PREDICTION METHOD (NORWAY) Two factors contributing to the good subjective quality demonstrated at the Paris meeting (AVC-30) are described are described: - multiple two-dimensional VLC's for DCT coefficient coding which have EOB of 1-3 bits and are switched adaptively for each coefficient without side information - dual frame prediction which uses only forward prediction with two preceding frames This scheme is characterized by low delay and simple architecture. AVC-105 TERMINOLOGY CHAPTER FOR THE STATUS REPORT (DIS, NCS) Definitions of relevant terms are provided for inclusion in 2 of the status report. Temporary Documents TD-1 Agenda for the third meeting in Santa Clara (Chairman) TD-2 Available documents (Chairman) TD-3 List of tape demonstrations (Chairman) TD-4 List of picture formats (Chairman) TD-5 Agreements and action points for the picture format issue (Small group on picture format) TD-6 Framework for discussion of H.26X format(Small group on picture format) TD-7 Additional submission materials for the "Kurihama Tests" (Experts Group) and a diagram illustrating N-ISDN and B-ISDN interworking (Chairman) TD-8 H.26X requirements (Small group on requirements) TD-9 Liaison statement to SGXVIII (Small group on SGXVIII liaison) TD-10 Liaison to TG CMTT/2 SRG (Liaison representative) TD-11 Draft report of the third meeting of the Experts Group for ATM Video Coding in Santa Clara (Chairman) END