FCD editors will consider the comments during editing FCD.
M15300
Summary of Voting on ISO/IEC CD 23002-4
FCD editors will consider the comments during editing FCD.
M15382
A fast simulation of RVC MPEG4 SP decoder using Cal2C code generation
Provision of Cal2C tool for fast simulation (or code generation)
Free software (BSD-like license)
Serves as a proof of concept
[Recommendation] make a short demo on Wednesday Plenary
[Recommendation] adopt the tool in RSM package. [RSM WD]
M15397
Swiss National Body comment on possible improvements of CD 23001-4 and CD 23002-4
Item 1: “that the informative specification of CAL language used for implementing the reference SW of the RVC Toolbox included in the 23001-4 CD is profiled to only the operators that admits direct synthesis to HW and SW implementations,” review of M15386
Item 2: “that ISO/IEC consider standardization of CAL language in the reduced form that admits synthesized SW and HW forms,” review of M15386
Item 3: “that CD 23002-4 updates the textual description of VLD and CA-VLD FUs including feedback tokens to the main parser FUs,” review of M15384
Item 4: review of M15389
M15451
Liaison on RVC
MPEG-B related CE
Doc. No.
Title
M15381
Result of RVC CE 1.1
[Recommendation] The CE 1.1 will be closed.
Issues discussed:
How frequent are the data updated? Not fully clear, depends on application
What is the size of a decoder description? Probably 2-3% of whole amount of data, depending on the frequency of retansmission.
Is it useful to transmit a decoder description 60 times per second as mentioned as an extreme case in the contribution? Considering the fact that decoder instantiation delay may even be longer, and also that I frames typically are only sent twice per second, having 2 descriptions per second transmitted seems more realistic
Hint by Olivier: Typically configuration parameters are sent offline (out of band) to settop boxes. Looking at decoder instantiatiation delays, the sending on-the-fly for real-time presentation seems to be useless.
This use case certainly does not have the highest priority right now. Further study needed, but this does not prohibit the other developments in RVC. Action could also be taken at a later point in time.
M15389
BSDL profile for RVC
Motivation: Inhibit various specifications of bitstream syntax for a given standard. Purpose: Simplify implementation of parser.
OK in principle. Will be put as annex in MPEG-B and denoted as “RVC-BSDL”
It must be possible to eventually extend it for future RVC systems (could e.g. be managed by version 1 or version 2 profiling)
The contribution provides a good example of common elements between different standards. It is shown that many FUs can be re-used between AVC and AVS (not the more important ones, however). Conclusion: RVC framework can fully support AVS.
It was also requested to mention the applicability of RVC supporting AVS in 21001-4 annex C. Indeed, it was concluded in further discussions that no actual mentioning of concrete standards should be put there, because we expect that any would be supported. It was thus decided to remove Annex C of MPEG-B part 4.
Issue a test report after finalization of RVC including the proof that it works with non-MPEG standards
M15418
BSDL Description of MPEG-2, MPEG-4 Simple Profile and AVC Bitstream Syntax for RVC Framework
BSDL schema specifying the usage of datatypes and variables provided and validated for MPEG-4 SP, MPEG-2 MP and AVC (BP?). Necessary for mapping of syntax elements into the associated operations of FUs. Context-adaptive entropy coding requires storage of temporal values in the schema. Validating of AVC bitstreams is not easy with the current BSDL reference software.
Is validation possible at all for context-adaptive schemes at the bitstream level? For example CABAC takes a binarized representation (coefficients, modes, vectors etc.) and then the coding engine further compresses them into binary, assuming certain contexts (decoding engine working inversely). Could this be seen as a two-stage approach with binarized representation expressed by BSDL, and CABAC decompressor as a kind of pre-processing unit that generates it?
Recommendation : include the specification in the FCD
MPEG-C related CE
Doc. No.
Title
M15387
On the application of the naming convention on Variable-Length Decoding FUs names
- the proposal has been approved and the new naming convention should be used in bot VTL and RSM. Christophe will apply the new convention to RSM and J-Shin will apply it to the FCD of MPEG-C.
M15388
Scheduling of CAL models for hardware/software implementation
- The contribution presents a procedure for the analysis of efficient scheduling for a RVC ADM. This work is a direct contribution to the development of CAL2SW and CAL2HW tools and should be part of the tool development workplan. Supporting tools should be included as informative parts of RSM standard or a separate informative TR. The recommendation is to join the CAL2C work and contribute to the development of such tools that will be an informative part of the RVC standard.
M15405
Textual descriptions of AVC inter perdition FU's for adaptive interlacing coding
- The contribution provides the textual description of new FUs for AVC inter prediction interlaced coding. The recommendation is to include the textual description in the FCD and upload the implementation in the RSM.
[Recommendation] will continue EE and put the outcome of the contribution to status report of EE
[Recommendation] Dandan presented the EE description and status report in the Video Plenary
Conformance testing
Doc. No.
Title
M15420
RVC Conformance Testing of a Functional Unit using Open Dataflow and TCP/IP
- The contribution proposes a TCP/IP implementation connecting reference SW to FUs testbeds for conformance testing. Recommendation: include the proposed methodology in the Conformance WD and the reference SW to be included as conformance tools in the RSM repository.
M15443
Procedure of validation of BSDL schema in RVC framework
Description of a procedure for the validation of schemas. For CAVLC and CABAC there is a problem: a full decoder needs to be implemented for validation or the syntax element of CAVLC and CABAC need to be ignored. However the “ignore” operator is not implemented in BSDL reference SW. proposed solutions is to describe in the FDC a “simplified BSDL validation procedure” when CAVLC and CABAC are not used and a complete validation procedure that implies that a full abstract decoder model is instantiated and implement a full decoding process that is used for validation of the BSDL schema.
Functions used in validation: BinToBsd and BsdToBin. Additional Javascript functions needed for right/left shift and log2 operations. It was discussed whether “ignore” could be used for certain bitstream sections to skip over CABAC-type encoded sections. For context-adaptive parts of encoded bitstream, it will be necessary to implement the real decoder as functional unit. The parser should work at a higher level here.
Conclusion: In principle, CAVLC and CABAC are implementable in RVC framework. CABAC implementation needs to be started urgently.
RSM
Doc. No.
Title
M15384
Automatic generation of parsers using XSLT and updated versions of the VLD FUs
the contribution describes minor changes to FUs for VLD needed to correctly synthesize a parser from a BSDL description. An update for the RVC toolbox is proposed.
Action point: update textual description of VLD and CAVLC tables and update RSM with CAL code.
M15385
On the synthesizability of the Video Tool Library in the RVC framework
[Recommendation] update the CAL guidelines & CAL specification to reflect the points in the contribution. (MPEG-C FCD)
M15386
CAL profile for HW/SW code generators of the RVC framework
[Recommendation] update the CAL guidelines & CAL specification to reflect the points in the contribution. (MPEG-C FCD)
M15388
Scheduling of CAL models for hardware/software implementation
