International organisation for standardisation organisation internationale de normalisation
Download 8.47 Mb.
|
Test Model ImprovementsIt was agreed that m30228 and m30309 will be integrated into the TM by 9th August 2013. Integration order: m30228 passed to proponent of m30309 who will merge and integrate. Anchors will be provided as usually by Visual Atoms and Telecom Italia (Linux) and Huawei and PKU (Windows) by 16th August 2013.
15.0.0.1.1.1.1.1.58m30480 CDVS: CABOX TM Implementation and Acceleration Notes [Gaurav Srivastava, Victor Fragoso-Rojas, Abhishek Nagar, Zhu Li] Implementation notes and details for m30466.
The proposed changes of m30309 will be implemented in the TM and WD documents. The proposed changes of m30228 will be implemented in the TM document. Three weeks editorial period is requested.
Two Core Experiments were defined: CE-1 on key-point detection and CE-2 on global descriptor.
The following issues were initially identified to be resolved during the meeting:
The following recordings of discussions were made in the BoG (chaired by Pierrick Philippe). 15.0.0.1.1.1.1.1.59m29689 VP8 Decoder Description [Mohamad Raad, Harald Alvestrand, Lazar Bivolarsky, Adrian Grange, John Luther] Presented by Mohamad Raad, Tuesday at 12:15 pm. Small editorial changes were made since the last meeting. Syntax and semantics are described, along with the decoding process. The proposer believes all the relevant components for implementing the technology are present in this description. A patent declaration is included at the end of the document; it is strictly the same as in m26691. No comments from the floor were made on the lack of completeness or maturity of this documentation. 15.0.0.1.1.1.1.1.60m29690 VP8 Encoder Technical Overview [Lazar Bivolarski, Mohamad Raad, Harald Alvestrand, John Luther] Presented Monday afternoon by Mohamad Raad. The CfP documents stated: “Quantization settings should be kept static. When change of quantization is used it shall be described.” The submitted bitstreams don’t keep the quantization steps static over frames, however the QPs are kept static inside a frame. In Section 2, it is explained how the QP values are selected for the frames. The initial QP value for a frame is selected based on the target bit rate and is refined to match the bit rate. It does not consider the distortion. Explanation is also given on the different kinds of frames and prediction modes. The code includes some functions written in assembly code and the corresponding C functions are also provided. 15.0.0.1.1.1.1.1.61m29691 Comment on VP8 Licensing Terms [Mohamad Raad, Harald Alvestrand, Lazar Bivolarski, John Luther] Presented Tuesday Morning by Mohamad Raad. This document present the licensing terms made by the proponents, cited below: Extracted from m29691: “Google, Inc. may have current or pending patent rights relating to the technology described in this contribution and, conditioned on reciprocity, is prepared to grant a Free of Charge license to an unrestricted number of applicants on a worldwide, non-discriminatory basis and under other reasonable terms for implementation of the resulting ISO/IEC International Standard (per box 1 of the ITU-T/ITU-R/ISO/IEC patent statement and licensing declaration form).” The above terms were extracted from the ISO/IEC licensing terms form for an ISO/IEC deliverable (see the complete form at http://isotc.iso.org/livelink/livelink/fetch/2000/2122/3770791/ITU_ISO_IEC_Patent_Statement_and_Licensing_Declaration_Form.pdf). The text of the type 1 declaration was copied from the ISO directives. Question: Does this licence say anything about the modifications of the technology? Is the licence provided conditional to the eventual modification of the technology in the course of CEs? Answer from the presenter: the wording used does not say anything about conditions. The group understands that Google provides a Type 1 licence for VP8 as-is, and for a modified version of VP8 made in the course of a standardization activity conducted by ISO/IEC. Google has identified some IPR owners and has come to an agreement (a cross-licensing agreement) with those companies in the course of the WebM project. This infomation is part of the m29693 document (in the comment section of the registration facility). Question: Do we have an indication that those IPR owners will provide a Type 1 licensing in the course of this project. The cross licensing agreement available at http://www.webmproject.org/cross-license/ was presented. From this document (in the “Licensed field of use” section) it appears that the cross licensing agreement is for VP8 without further modification. From the discussion it can be concluded that any modification applied on the proposed technology would not be covered by the Cross Licensing Agreement. Related to this contribution, the Finnish National Body comment (m30508) was also presented during this session. Point 3 of the FINNB comment, related to the licence and software copyright was addressed. The copyright included with the software provided by the VP8 proponent was reviewed. The source code is governed by a BSD-style licence that can be found in the LICENSE file. (provided in the presentation attached to document m29693) A side-by-side comparision of the ISO/IEC license, as included in the HEVC code, and the VP8 license is performed. The differences are highlighted, only the owner is different (Google rather than ISO/IEC) the other terms are strictly identical. Question about the Copyright owner: should the software insert a copyright from the ISO/IEC: so far Google is the only contributor, hence having Google seems relevant at this stage. The Software patent text (included with the code) was also reviewed. The second half of the text says that limitation on the use of the VP8 software can be set in case of patent ligitation. Question for the proponent whether this text will be removed from the software patent in the case of this software would become the reference software. The proponents believe that this part would not be included in the potential reference software. Possible options for this proposal were listed, they are summarized in the table below:
Table of possible options for the VP8 submissions Options 1 and 3 are compatible with the IPR statements provided by the technology providers. As far as the native VP8 mode is available, this native mode is covered by the licensing terms provided by Google and in the Cross Licensing Agreement. 15.0.0.1.1.1.1.1.62m30539 Regarding changes to the VP8 codec [Mohamad Raad, Harald Alvestrand, Lazar Bivolarski] Presented by Mohamad Raad, Tuesday at 12:30. The documents state that the licensing situation is secure if no changes are made to the proposed technology. This document supports option 1 as listed in the table above and discourages option 2 and option 3. 15.0.0.1.1.1.1.1.63m29692 Regarding the potential CE process for the Internet Video Technologies activity [Mohamad Raad, Harald Alvestrand, Lazar Bivolarski, John Luther] Covered in a joint BoG with the IVC ITM development group. (Tuesday 2PM) 15.0.0.1.1.1.1.1.64m29693 Google Inc.’s response to the CfP on Internet Video Technologies [Harald Alvestrand, Adrian Grange, John Luther, Lazar Bivolarski, Mohamad Raad] The summary of the response to the call for proposal was presented by Mohamad Raad, during the AHG and during this BoG (Monday at 3pm). The document was updated and now includes a presentation giving an overview of the presentation (in the comment section of the WG11 input document repository). VP8 is presented as a mature technology, already vastly used in the industry. BD Rate performance (based on the HEVC anchors) are briefly presented. It was pointed out that the latest AVC anchors (based on JM18.2 and new configuration files) were not available for the proponents. The proponents clarified the situation of VP8 wrt to standardization. VP8 is not standardized elsewhere. A VP8 specification exists as IETF RFC 6386; however VP8 was never proposed as a standard to the IETF. IETF RTCWeb refers to the VP8 technology and is not an IETF standard. Recommendations from the proponents:
The submitted material was examined to identify the bit rate fulfillment. The excel spreadheet provided along with the document was presented, highlithing the bit rates of the provided coded items. There were overshoots on the bit rates compared to the CfP target bit rates
The following graphs present the histogram of the bit rate deviation, as % of the CfP bit rates for CS1 and CS2 from which the above conclusions were derived. 
15.0.0.1.1.1.1.1.65Some remarks on the subjective evaluation A MOS score of 3 means fair quality. It is noted that 14 of 19 tested items have above fair quality from the middle rate. The scores spanned by the VP8 show a good level of quality for the upper range of the CfP bit rates Further details are provided in the output document describing the result of the subjective testing (N13775). 15.0.0.1.1.1.1.1.66m30542 Description of AVC anchor generation for the IVC CfP [Rickard Sjöberg, Kenneth Andersson (Ericsson)] This contribution was covered during the AhG on CfP. Details can be found in the AhG report (m30534). 15.0.0.1.1.1.1.1.67Session on the preparation of the WD A session for the preparation of a WD text for the proposed CfP technology was held. The technology description (document m26689) was reshaped into a WD document (N13776). The following points were identified:
Further issues:
|