Iso/iec jtc 1/sc 29 N
Yüklə 144,65 Kb.
|
|
INTERNATIONAL ORGANISATION FOR STANDARDISATION ORGANISATION INTERNATIONALE DE NORMALISATION ISO/IEC JTC1/SC29/WG11 CODING OF MOVING PICTURES AND AUDIO ISO/IEC JTC1/SC29/WG11 MPEG2009/N10714 July 2009, London, UK Source: Video Subgroup Status: Approved Title: Text of ISO/IEC 23002-4/FPDAM 1 Video Tool Library Conformance and Reference Software Editor: Gwo Giun (Chris) Lee, Christophe Lucarz, and He-Yuan Lin ISO/IEC JTC 1/SC 29 Date: 2009-09-9 ISO/IEC 23002-4:200X/FPDAM 1 ISO/IEC JTC 1/SC 29/WG 11 Secretariat: Information technology — MPEG video technologies — Part 4: Video tool librar, AMENDMENT 1: Video tool library conformance and reference software Élément introductif — Élément central — Partie 4: Titre de la partie
This document is not an ISO International Standard. It is distributed for review and comment. It is subject to change without notice and may not be referred to as an International Standard. Recipients of this draft are invited to submit, with their comments, notification of any relevant patent rights of which they are aware and to provide supporting documentation.
This ISO document is a Draft International Standard and is copyright-protected by ISO. Except as permitted under the applicable laws of the user's country, neither this ISO draft nor any extract from it may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, photocopying, recording or otherwise, without prior written permission being secured. Requests for permission to reproduce should be addressed to either ISO at the address below or ISO's member body in the country of the requester. ISO copyright office Case postale 56 CH-1211 Geneva 20 Tel. + 41 22 749 01 11 Fax + 41 22 749 09 47 E-mail copyright@iso.org Web www.iso.org Reproduction may be subject to royalty payments or a licensing agreement. Violators may be prosecuted. Foreword ISO (the International Organization for Standardization) and IEC (the International Electrotechnical Commission) form the specialized system for worldwide standardization. National bodies that are members of ISO or IEC participate in the development of International Standards through technical committees established by the respective organization to deal with particular fields of technical activity. ISO and IEC technical committees collaborate in fields of mutual interest. Other international organizations, governmental and non-governmental, in liaison with ISO and IEC, also take part in the work. In the field of information technology, ISO and IEC have established a joint technical committee, ISO/IEC JTC 1. International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2. The main task of the joint technical committee is to prepare International Standards. Draft International Standards adopted by the joint technical committee are circulated to national bodies for voting. Publication as an International Standard requires approval by at least 75 % of the national bodies casting a vote. Attention is drawn to the possibility that some of the elements of this document may be the subject of patent rights. ISO and IEC shall not be held responsible for identifying any or all such patent rights. Amendment 1 to ISO/IEC 23002 4:200X was prepared by Joint Technical Committee ISO/IEC JTC 1, Information technology, Subcommittee SC 29, Coding of audio, picture, multimedia and hypermedia information. Information technology — MPEG video technologies — Part 4: Video tool librar, AMENDMENT 1: Video tool library conformance and reference software
After 3.11, add the following:decoder configurationconceptual configuration of a decoding solution.
In summary a RVC decoder description essentially consists of a list of FU and of the specification of the FU connections (FND expressed in FNL) plus the implicit specification of the parser in terms of bistream syntax description (BSPD expressed in RVC-BSDL). So as to be complete behavioural model (the ADM) an RVC decoder description needs to make reference to the behaviour of each FU that is provided in terms of I/O behaviour by the RVC video toolbox library specified in RVC-CAL. model instantiationbuilding of the abstract decoder model from the decoder description (consisting of the BSD and the FND) and from FUs from VTL NOTE During the model instantiation, the parser FU is reconfigured according to the btstream syntax description (BSD) or loaded from VTL. FU network languageFNLlanguage that describes the FND, known also as “network of FUs”. BSDLdescription of the bitstream syntax and the parsing rules NOTE Bitstream Syntax Description Language (BSDL) was standardized in ISO/IEC 23001-5. RVC-BSDLpertinent subset of BSDL which is defined within the scope of the current RVC framework Black box testingtesting that specifies only the input, output of the object under verification, where the internals of the object being tested are irrelevant and therefore hidden. Add the following after 7.3:8Conformance testing for MPEG-C Part 4 (ISO/IEC 23002-4)In order to warrant the reusability, maintainability, and flexibility of the functional units, their corresponding testbeds should also be created for conformance validation prior to inclusion in the RVC specification. These testbeds also serve as executable “application notes” which provide the users with concise procedures on the utilization of the RVC video tool library. Furthermore, specification of the I/O behaviors of the RVC video tool library is also achieved by executing the tool with its conformance testbed. Hence, the testbed should not only verify the functionality but also validate the conformance of the FU’s. 8.1Internal description of FU’s and Conformance Testing EnvironmentAs mentioned in previous sections, the RVC toolbox is composed of functional units and tools to test the conformance of functional units as shown in Figure AMD1.1 [2]. 
Figure AMD1.1 — Basic Functional Unit of Video Tool Library FU’s in RVC are modeled as actors in the CAL language, where the behavior of the FU’s depend on the input tokens and the internal state of the actors. Benefiting from the CAL-based FU’s, the verification processes of the FU’s should be performed by feeding tokens to the FU’s and checking the state and corresponding output tokens, since the functional units are encapsulated independently. The conformance testing tools should also be implemented in CAL, where the testing tools act as the testbed for functional units in Moses as shown in Figure AMD1.2 . The testbed is composed of the test pattern database, the golden response database, and the automatic comparison and analysis unit. The test pattern database contains the test patterns or tokens to feed the functional unit under verification or testing for different purposes. The tokens are categorized (tbd) into different classes and the tokens in both the test pattern database and the expected golden response database should be conformal to the textual definitions in the RVC specification. Scheduling of the tokens will also be described in the textual description for conformal testing purposes (tbd). 
Figure AMD1.2 — Conformance testbed of functional units 8.1.1Generation of the Test Pattern and Golden ResponseTest patterns or predefined input tokens and the corresponding golden responses or expected output tokens to and from the FU’s have been generated from the corresponding MPEG reference software by using the associated development kit from CAL (tbd). 8.1.2ConformanceThe verification tests should be conformal to the normative textual description for the corresponding functional units in the RVC Video Tool Library specification. For conformance purposes, all normative portions of the FU’s should be verified using black box strategy. 8.1.3Different categories of verification TestsThe verification tests in RVC conformance tools are categorized into FU level and CODEC level testing. During the addition of new beneficial coding tools to the tool library in MPEG RMC/RVC, the corresponding conformal tests should also be specified and performed before being accepted in the ISO/IEC 23001-4 and ISO/IEC 23002-4 specifications. In addition, systematic top down procedures for codec level conformance testing should also be generated. 8.1.4Verification test coverageCoverage analysis as described below is used in the informative verification of RVC to measure the quality of the testbed quantitatively. This analysis will indicate both the parts of the functional units that have been executed and also specify places where further verification are required. The major types of coverage measures include action coverage, condition coverage, branch coverage, and path coverage. 8.1.4.1Action coverage: The percentage of all the statements of actions which have been exercised in the actors of a FU for CAL codes.8.1.4.2Branch coverage: The percentage of the branches which has been tested, where the branch is the outcome of a decision.8.1.4.3Condition coverage: The percentage of the outcome (true or false) of each conditional sub-expression that has been executed.8.1.4.4Path coverage: The percentage of the paths in the FU that has been executed. A path is the independent sequence of all the possible combinations of the branches in the FU.8.2Description of Functional Unit Conformance Testing8.2.1Test ProcedureThe functional units of RVC are classified into two major categories, namely algorithmic content FU’s and data management FU’s [7]. 8.2.1.1Algorithmic content FU’s: This type of FU comprises a coding tool of a specific MPEG standard with particular profile-and-level. The behaviors of an algorithmic content FU should be completely conformal to the correspding specifaction described in MPEG standard document.8.2.1.2Data management FU’s: This category of FU packs the required decoding information and data into tokens in CAL and feeds the tokens into the other functional units in the correct order.The conformance test requirements for these two types of functional units are slightly different, since the functionalities of algorithmic content FU’s have been specified in the existing MPEG standard but the data management FU’s, however, are brand-new and under development. Therefore, for the data management FU’s, it is mandatory that test patterns and the corresponding golden response achieve 100% test coverage to guarantee the functionality of such FU’s. In order to claim that an algorithmic content FU is compliant to a specific MPEG standard with particular profile and level, the FU is strictly required to successfully pass all the conformance test cases extracted from decoding the existing conformance bitstreams of the normative test suite by using the corresponding reference software. Note that normative test suites of each standard can be found in the electronic annex of the MPEG standards. In addition, the testbed for the algorithmic content FU’s must accomplish 100% test coverage as well. After RVC is standardized, some of the test patterns and golden response generated for verification of functional units will serve as the conformance test suites of FU’s directly. In order for the FU’s to be included in the RVC specification, it is mandatory that at least one successful test is performed from every coverage categories including action coverage, condition coverage, branch coverage, and path coverage (TBD). 8.2.1.3Conformance testing procedures for ALGO FUThe functionalities of ALGO FU’s have to be fully verified with the following procedures: Verify all the modes of operations for the ALGO FU with
i. the token granularity used and tested, ii. the parameter settings, etc.
8.2.1.4Conformance testing procedures for MGNT FUsThe conformance testing of MGNT FUs is conducted when conformance testing of network of FUs is performed. 8.2.1.5Conformance testing procedures for network of FUsThe network of FUs in this context is tested using the same procedures as described in 8.2.1.3 for the ALGO FUs. This time the network of FUs can be considered as a virtual ALGO FU in that it receives input tokens and produces output tokens. A network of FUs may contain a number of ALGO FUs as well as MGNT FUs. A network of FUs may also contain several networks of FUs 8.2.2List of functional units for video tool library and associated conformance testing8.2.2.1Syntax Parsing
8.2.2.2Syntax Parsing FU for MPEG-4 Simple Profile
8.2.2.3Texture Decoding FU for MPEG-4 Simple Profile
8.2.2.4Motion Compensation FU for MPEG-4 Simple Profile
8.2.2.5Syntax Parsing FU for MPEG-4 AVC Constrained Baseline Profile
8.2.2.6Texture Decoding FU for MPEG-4 AVC Constrained Baseline Profile
8.2.2.7Motion Compensation FU for MPEG-4 AVC Constrained Baseline Profile
8.3Conformance Testing at Decoder Level8.3.1Test Procedure8.3.1.1Conformance testing procedure for a decoderA decoder can be considered as a network of FUs that contains all the necessary FUs and networks of FUs inside. Therefore, the conformance testing of a decoder can be conducted in the same manner as in 8.2.1.5. The only difference is that the input token of the decoder network is the conformant bitstream. 8.3.1.2Conformance testing and acceptance procedures for a new decoderFor accepting a new decoder in RVC framework, it is expected that the proposers bring the conformance testing of newly described FUs and network of FUs that contains the newly described FUs. If the new decoder is drawn from the supported decoder in RVC, the proposers can skip the conformance testing of the existing FUs and network of FUs in RVC conformance test. To test the newly described FUs and network of FUs, the proposers should follow the procedures as described below:
8.3.2MPEG-4 SP8.3.3AVC BP9RVC Simulation Model (RSM)9.1OverviewThis Clause describes the RVC Simulation Model. The following figure illustrates the RVC platform. 
The Decoder Description is composed of:
The encoded video data is a binary bitstream composed of 0s and 1s. [Binary file] The Abstract Decoder Model (ADM) is the instantiated decoder model. It is composed of the FUs and their connections. The ADM is written in CAL (for actors) and in XDF (for the networks). [Support: Open DataFlow environment] The Video Toolbox Library (VTL) is a collection of FUs written in CAL and XDF. The Model Instantiation comprises:
The Decoder Implementation is the methodology of the designer to implement the ADM into implementation languages. CAL2C and CAL2HDL tools are under development to transform the ADM into an implementation representation (C and VHDL for example). During this process, the designers can pick FUs which are implemented in a target language and belong to the proprietary toolbox. [Support: code generators (CAL2C and CAL2HDL) under Open DataFlow environment] The Decoding Solution may be any final implementation of the decoder. For example, the MPEG-4 Simple Profile decoder has been completely translated into hardware on a FPGA, this is a decoding solution that fills some requirements. The decoder video data is the video itself, outputted in a given format.
9.2Guidelines to build a RVC platformConstructing a new decoder requires modifying at least one of the following items:
The different steps of the procedure to build a new decoder with the RSM are: 9.2.1Encoder side issuesDefine the bitstream The bitstream can be build with a C/C++ reference software or using any methodology that enables to output a binary file. The equivalent RVC-BSDL description of the bitstream must be done/completed by the designer himself. The outputs of this task is the RVC-BSDL description of the bitstream and the bitstream itself in a binary format. Build the FU network description Once all the FUs are designed, they must be connected together to build the decoder. The designer can pick in the VTL to complete his decoder with existing FUs. It consists simply to copy paste the necessary FUs into the working directory under OpenDF. Thus, the output of this task is the Decoder Description (in FNL). 9.2.2Video Tool Library issueBuilding the FUs For a designer to test his new decoder, he needs to write his new coding tools in CAL and to integrate it into VTL. The designer will design his FU under the OpenDF environment. The outputs of this task are the FUs written in CAL necessary for the new decoder. 9.2.3Decoder Composition MechanismBuilding the parser FU The generation of the main parser actor has been automated. It generates a parser written in CAL and FNL from the bitstream schema (in RVC-BSDL) describing the bitstream. This transformation process is written in XSLT. The generated parser FU is a hierarchical FU: it contains a main component and several other components connected together in order to parse the bitstream (e.g. VLD tables).
Pre-processing is the first operation conducted by the top level stylesheet. The pre-processing collects the individual schemata into a single intermediate tree, taking care to correctly manage the namespace of each component Schema and also performs a number of other tasks, including assigning names to anonymous types and structures. The processing is divided in the main CALML subparts. Globals includes all constant/variables used in the actions. Actions provide all the operations with their guard, input, output and behavior. Transitions define the state machine passing from one state to another by the actions. Priorities define the priorities between different possible actions available at a same state. CALML Templates collects templates used by other modules to generate the output. Finally CALML Structure put all the things together to generate the new CALML Parser Actor. The output of this task is the CAL and FNL codes describing the hierarchical parser FU built from the BSDL schema of the bitstream.
This task consists in connecting the other necessary components with the main component, using NL. The output of this task is the FNL showing the connection between the main parser actor and its parsing FU (for example VLD FUs). Building the Abstract Decoder Model It consists in picking the FUs from the VTL and to assemble them according to the FU network description. Designer copy the files of the FUs (.cal files) and insert them into the OpenDF simulator. At this step, the ADM design which is composed of CAL actors and their respective connections has been successfully output based on the Decoder Composition Mechanism.
Implementation begins when the ADM is available. This step can be done by several ways and it is not normative. Designers can use whatever methodology to implement this ADM. For the moment, some tools already exist to generate automatically C and Verilog/VHDL code from the ADM. The generated code can lead to an implementation. Table of mapping software modules with the FUs (new section) The Decoding Solution At this end, the designer obtains a decoding solution, which is the means to decode a bitstream to produce video data. Bibliography
Document type: Document subtype: Document stage: Document language: Yüklə 144,65 Kb. Dostları ilə paylaş:
|