International organisation for standardisation organisation internationale de normalisation
Yüklə 7,43 Mb.
|
- Bu səhifədəki naviqasiya:
- New CE definition
- Web Video Coding
- Internet Video Coding Exploration
- Review of documents
WD and TM Development1.1.1.1.1.29m33194 CDVS: TM9 Extraction Evaluation on ARM Architectures [Giovanni Ballocca, Alessandra Mosca, Attilio Fiandrotti, Massimo Mattelliano] TM9 release 166 was tested using CDVS building dataset + ARM v6 and v6, Device LG P500, LG Nexus 5 Server 64 core AMD, 128GB RAM, Ubunte 2012.10 3.2.0. MAP retrieval: 0.5KB – 56% MAP, 16kB - 69% Extraction: ARMv6 -11s, ARM v7 – 2s. Some strange behaviour was observed for extraction time at 16k - -.25s. It was recommended that a power profiler be used and for results of this to be reported.
CE 1 was define in preparation for Conformance testing & to perform a general check of the performance and stability at the MPEG 108 after changes are introduced.
TM10 schedule:
ISO/IEC DIS 14496-19 (N13561) had been issued in Incheon. Resolutions had been formulated by the same time and renewed in Vienna, urging for action on submitting IPR statements before the closing of the DIS ballot. Since then, the situation has slightly improved, even though some information in the data base still appears to be incomplete. Nevertheless, in order to make progress, the FDIS was issued as N14387 (and DoC N14386). In response to NB comments, it had to be considered that decoder conformance in Annex C is expressed in a manner dependent on the buffering period SEI message and the picture timing SEI message. Also some normative expressions relating to SEI messages are found elsewhere – e.g., Annex E. Therefore, a normative reference to AVC was needed. 1.1.1.1.1.30m33245 Comment on the WebVC project [David Benham]
The BoG on IVC was chaired by Ronggang Wang. The first meeting of the BoG was held Monday afternoon. The following issues were initially identified to be resolved during the meeting:
The subsequent notes on documents were recorded by Ronggang Wang, partially from the AHG report, where an initial review was performed . As far as the BoG made recommendations, these were later approved by the video plenary in the context of output documents (adoptions to ITM N14429, WD N14428 and Prior Art documentation N14431).
The following input documents had initially been reviewed in the AHG 14-03-30 Meeting and further discussions were performed in the BoG during the meeting week. 1.1.1.1.1.31m32727 Improvement on Non-Reference P frame coding for IVC [Dong-Hyun Kim, Jae-Gon Kim] Abstract: An adaptive non-reference P frame coding with 3-level coding structure is proposed to replace current 2-level structure, average performance gain of 0.2% and 0.7% is reported in CS1 and CS2 respectively. Discussions: Prior art? Based on the earlier contributions of IVC. Adaptive hierarchical coding can give about 1.3% bitrate saving compared with fixed hierarchical structure. Conclusion: It was recommended to give more information on prior art on adaptive hierarchical coding. It was later confirmed that prior art of 1987 exists on this aspect. (Patent US 5699476, filed in 1987, notice in ISO patent database that it has expired in 2010. No flexible referencing mechanisms, static structure.) 1.1.1.1.1.32m33148 Cross-check of Improvement on Non-Reference P frame coding for IVC (M32727) [Minsoo Park, Myeong Kyun Kim, Sang-hyo Park, Euee S. Jang] Abstract: There is no mismatch of overall BD-rate between the verification results and the proponent’s results. 1.1.1.1.1.33m33144 Updated Internet Video Coding Test Model (ITM) v 8.0 [Sang-hyo Park, Jin Yeon Choi, Seung-ho Lee, Myeong Kyun Kim, Minsoo Park, Euee S. Jang] Abstract: An updated IVC test model, especially on the encoder side is provided by this contribution. Added new coding tools since ITM3.0. In addition, some problems on IVC WD1.0 are also reported. For example, the 8x8 transform should be improved to resolve the mismatch between WD and software. Discussions: Add new coding tools since ITM3.0 to Test Model. Where the hierarchical coding structure stated? On the frame-level encoding. Revise the inverse transform implementation in software to match the WD document. Conclusion: Recommend to adopt the two comments from this contribution. Recommend to adopt the new skeleton of IVC Test model. 1.1.1.1.1.34m33146 Extended Deblocking Method for Internet Video Coding [Myeong Kyun Kim, Minsoo Park, Sang-hyo Park, Jin Yeon Choi, Euee S. Jang] Abstract: Extend the curve fitting de-blocking method proposed in previous meeting to support adaptive blocksize transform adopted in ITM8.0, subjective quality improvement in low-rate cases are reported. Discussions: The method is only applied on I frame? No. Was the video quality improvement checked? It is very hard to find the difference between anchor and proposal. Although there is no increase in decoding time, But inverse matrix implements issue in chip design? Try to avoid inverse matrix operation if it is an issue. It is applied in very limited cases. We may not be sure the subjective quality improvement in all cases. The current implementation is not good enough for subjective quality improvement. Conclusion: It was recommended to do expert viewing during this meeting on some sequences to find the subjective quality improvement. It was later confirmed that no subjective improvement could be found. The technique goes back to a publication of Deknuyd et al. in 1994. Deblocking for Intra based on DC prediction, had already been used for 8x8. Seems OK. Some more assessment of visual benefit is planned before adoption. 1.1.1.1.1.35m33140 Chroma interpolation filter for internet video coding [Zhenyu Wang, Hao Lv, Xufeng Li, Ronggang Wang, Siwei Ma, Tiejun Huang, Wen Gao] Abstract: A 4-tap interpolation filter is proposed to replace current bi-linear interpolation for chroma component, the average performance gain for Y,U and V is 1.9%, 5.9%, and 5.8% respectively in CS2, and 0.2% , 0.6% and 1% respectively in CS1. Discussions: Q: is the same as 1997’s tech? A: not exactly. Reused from previous. 1997 do not mention luma or chroma. A: We have Patent for coefficients Q: Don’t change coefficients by resolution? A: yes Q: Why don’t use multiple filters as luma component A: complexity issue Q: Did you check complexity issue compared to bi-linear? A: For encoder only. Decoder can be increased. Q: Bitrate improvement is meaningful. Conclusion: Recommend to adopt. For informative information, Decoding and encoding time increase will be checked till during this meeting. Filter coefficients go back to Lanczos 1979. A patent application on the specific coefficients used was filed prior to the last meeting by the proponents, but not granted yet. It is expressed by the proponents that they would be willing to apply type-1 licensing to the patent if it would be granted. Follow-up action may be necessary if the patent is not granted due to prior art that is not type-1. 1.1.1.1.1.36m33313 Cross-check of M33140 on chroma interpolation [Weiran Li, Yun He] Abstract: This contribution provides cross-check results of a chroma interpolation coding tool in IVC proposed in m33140. The software provided by the proponent was reviewed and is consistent with the descriptions in m33140. The coding efficiency of cross-check is identical to the ones reported by proponent. 1.1.1.1.1.37m33381 Cross-check of Extended Deblocking Method for IVC (m33146) [Dong-Hyun Kim, Jae-Gon Kim] Not available yet during initial review of m33146. Was presented later. Generally, information about type-1 licensing situation seemed more convincing than in past IVC proposals.
|