International organisation for standardisation organisation internationale de normalisation

7.9Test Sequences

See also notes relating to JVT-Y072.

7.9.1.1.1JVT-Y077 ( Info) [C. K. Chung, H. Song, J. H. Park, B. H. Choi (KETI)] Multi-view test sequence for MVC

KETI has prepared one multi-view test sequence that is being made publicly available for standardization and research efforts in this area. Further details on these sequences were provided in this document.
At the Geneva meeting, KETI showed the multi-view test sequence but reportedly did not provide enough backup data such as camera parameters and comparison between JM and JMVM.
At this meeting, KETI has prepared other multi-view video sequences which have 8 views, and the size of these sequences is 640x480x300 frames.
The capture system is described in JVT-X082 and there is no change to it. Please refer to JVT-X082 for further detail on that topic.
No restrictions on the use of these sequences were requested.
The sequences can be accessed directly with the following ftp information

FTP address: ftp://203.253.130.48

Username: mvc

Password: KETI (Case-sensitive)

Directory: /MVC/Sequence
KETI reportedly compared Rate-distortion curve between JM 12.4 and JMVM 5.0. As shown Fig.2 of the contribution, JMVM 5.0 reportedly always shows better coding performance than that of the JM for all points. JMVM tools (IC 2-3% and MS <1%) reportedly show better coding performance, but the degree of performance improvement seems to be not so significant.
It was suggested to add these sequences into our common conditions. That topic is deferred for further study.

7.9.1.1.2JVT-Y079-V [Y. S. Ho, K. J. Oh, C. Lee (GIST)] Verif JVT-Y077 MVC test sequence? <>

8New AVC Proposals

8.1Adaptive chroma sampling

8.1.1.1.1JVT-Y034 ( Prop 2.2/3.1) [S. Sekiguchi, S. Yamagishi, Y. Yamada, K. Asai, T. Murakami (MEI)] On adaptive chroma sampling for 4:4:4

(Presented jointly with MPEG video.)

In this contribution, the result of an investigation of the value of adaptive chroma sampling mechanism for further performance improvement of 4:4:4 video coding was reported. In the last (Geneva) meeting, JVT-X044 asserted that coding performance of High 4:4:4 predictive profile provides a reasonable rate-distortion trade-off with improved color reproduction compared with conventional 4:2:0 coding. It also asserted that YUV based coding performance is generally superior to that of direct RGB coding in very low bit rate conditions, however, the reconstruction quality of luma signal can reportedly be degraded due to using more bits for the additional chroma samples. Based on this observation, a preliminary estimation of the value of adaptive use of 4:2:0 chroma sampling within 4:4:4 coding has reportedly been made by collecting macroblocks having better rate-distortion trade-off from 4:2:0 coded results vs. 4:4:4 coded results. Even though it does not take into account the difference of prediction reference fidelity, it was suggested that this adaptivity could improve luma reproduction fidelity while maintaining that of chroma components.
Macroblock-adaptive 4:4:4 / 4:2:0 coding was proposed (examples were shown where color disappears in very small colored areas). Results were shown only with Y-PSNR, which does not tell very much (worse performance than 4:2:0 coding, but better than 4:4:4 coding). Currently, no evidence appears to have been provided about the real benefits. Further study is needed, also in context of potential complexity increase by macroblock-wise switching as compared to the benefit that it may give.

8.2New levels

8.2.1.1.1JVT-Y035 ( Prop levels 2.0/3.1) [Y. Bandoh, S. Takamura, Y. Yashima (NTT), Shinichi Sakaida, Yukihiro Nishida (NHK), T. Itoh, A. Nakagawa (Fujitsu), I. Ando, H. Takehara (JVC)] Prop new higher levels

(Corresponding WG 11 contribution M14890.)
This document proposes new levels for MPEG-4 AVC/ITU-T H.264 beyond the current level 5.1 in order to support higher resolution video formats such as 3840x2160/60p and 7680x4320/60p.
Higher resolution video format systems beyond HDTV called UHDTV have reportedly been developed and are reportedly emerging in the industry. These formats have reportedly already been standardized in ITU-R Recommendation BT.1769. For the transport of these video signals, ITU-T Recommendation J.601 identifies MPEG-4 AVC/ITU-T H.264 as one of coding schemes, noting the necessity of new higher levels. Codec systems based on MPEG-2 and MPEG-4 AVC/ITU-T H.264 for 7680x4320/60p have reportedly been developed using multiple HDTV codec units by dividing the complete images spatially and temporally. Some transmission experiments via gigabit IP networks were reportedly demonstrated at NAB2006 and IBC2006, and an experimental live relay broadcast from Tokyo to Osaka for public viewing was also conducted using a commercial IP network.
MPEG-4 AVC/ITU-T H.264 currently specifies a maximum level known as level 5.1, which supports a maximum frame size of 36 864 macroblocks and a maximum macroblock rate of 983 040 macroblocks/sec at up to 240 Mbit/s. This corresponds to a maximum frame rate of 30.34 frames per second for 3840x2160 and 7.5 per second for 7680x4320.
It was proposed to specify new higher levels 5.2, 6, and 6.1 in MPEG-4 AVC/ITU-T H.264 corresponding to the following scenarios:

Level 5.2: 3840x2160/60fps (up to 4K x 2K / 60fps) 400 Mbit/s – 5 ref frames

Level 6: 7680x4320/30fps (up to 8K x 4K / 30fps) 960 Mbit/s – 5 ref frames

Level 6.1: 7680x4320/60fps (up to 8K x 4K / 60fps) 1600 Mbit/s – 5 ref frames

Current consideration is reportedly being conducted toward introduction of higher resolution formats in 2011 and beyond.
Remarks: AVC already has 3 levels that nobody has implemented yet. This could also be done by running different decoders in parallel.
Question: Are these bit rates realistic? Think about it.
Remark: 5 reference frame requirement may also be a little high.
Remark: Consider slice structure constraints, etc.
Remark: Consider >= 72 fps.
JVT & MPEG are interested to hear more about these developments, but currently it seems to be too early to go for a concrete definition of new levels. MPEG will draft a response to the Japanese NB on this topic.

8.2.1.1.2JVT-Y086-L JNB input on additional levels of AVC and related subjects

(Corresponding WG 11 contributions: M14863, M14868, M14869, and M14890.)
Presented jointly with MPEG video.
Summary: 1920x1080 up to 120/240p, 4kx2k/60p, 8kx4k/60p is being prepared/discussed in industry and standardization bodies such as SMPTE. It should be avoided that proprietary definitions appear in the market, and it was therefore asserted that MPEG should start consideration of defining additional profiles rather soon. 1920x1080 at 60p will reportedly happen very soon, and next will reportedly be 4kx2k. Higher resolution formats are still farther away.
Comment: We do not usually direct the definition of video formats, but try to accommodate as many formats as possible that are common in the market (or are soon expected) with a given level. In cases where formats are just emerging, it might be premature to make definitions. Definitions of maximum macroblock numbers, maximum buffer sizes, maximum bit rates are also important to consider when a level is defined and typically require some more careful study. Some of these values may also be different for consumer or professional applications.
Another option for implementation of larger formats would be, e.g., to combine four 4Kx2K tiles into on 8Kx4K frame (it is also an issue of decoder complexity whether a high-resolution profile is doable, could be done at systems level).

8.2.1.1.3General discussion of AVC new levels issues

Remark: AVC already has 3 levels that haven’t been implemented yet.
Remark: Could also be done by running different decoders in parallel.
Applications? Broadcast starting in 2011.
Plan to study for future definition –there is time to study.
MPEG & JVT are interested to hear more about these developments, but currently it is too early to go for a concrete definition of new levels. These statements are to be considered in the drafting of a response to the Japanese NB by WG 11.
JVT response: We plan to provide new levels in the future as necessary to serve industry deployments of important services; however we plan to avoid doing this prematurely and to ensure thorough and adequate study in creating such definitions. No immediate action – further study encouraged.