International organisation for standardisation

No.	Title	TBP
Available
11095	Description of Exploration Experiments in 3D Video Coding	N	10/01/22
11201	Draft Report on Experimental Framework in 3D Video Coding	N	10/01/22

6Explorations – High-Performance Video Coding

The main focus on the subject of High-Performance Video Coding was the finalization of the Call for Proposals. Terms of Reference (N11112) had been negotiated between the management of the parent bodies of the upcoming Joint Collaborative Team (JCT) on video coding, and were reviewed and approved. Accordingly, the preparation of the final CfP text was conducted jointly with VCEG. There were no substantial problems identified in the previous draft CfP text; however considering that 31 pre-registrations had been received, and that the testing relating to the CfP is expected to become a major effort, careful consideration of the overall timeline was made, and in particular the issue of payment for the testing fee was clarified (see responses to NBs, N11115). The final registration declaration period was opened immediately after the Kyoto meeting.

As the previous version of the document on Vision, Applications and Requirements of High Performance Video Coding (as of Lausanne, February 2009) was partially incomplete and outdated, a new version (N11096) was issued to accompany the CfP. It clarifies numerous aspects in the vision section, with higher compression as main target, however broadening the list of possible applications, e.g.

• 
  Home cinema, Digital Cinema, Broadcast
  
• 
  Surveillance
  
• 
  Real-time communications, video chat, video conferencing
  
• 
  Mobile streaming, broadcast and communications
  
• 
  Storage, Camcorders
  
• 
  Video On Demand, Internet streaming, download and play
  
• 
  3D video, telepresence
  
• 
  Content production and distribution
  
• 
  Medical imaging
  

The requirements section was updated – mainly to remove unclear and vague descriptions for various aspects, mostly without changing the substance. Some non-mandatory (“should”) requirements were added, e.g. support for lossless, support for alpha channels. Unnecessary restrictions on frame rate were removed. Some discussion was held on the issue of potential needs for support of interlaced-scan video; however no significant change was made to the requirements document in that regard . HVC remains intended to provide mandatory support only for progressive scan. For interlaced-scan video, it remains unclear whether there will need to be specific support in the scope of HVC applications, and if so, whether the available tools for progressive-scan encoding cannot be adequately used for interlaced-scan video with sufficiently high coding efficiency (e.g. using field coding).

Several contributions had been submitted with descriptions of specific algorithmic technology features that may be useful in the future work on the project (M17163, M17180, M17181). These were presented and discussed, and the participants were suggested to continue the study of such design aspects toward progress in preparation of CfP responses and for the subsequent work on the project.
For a smooth start of the collaboration with ITU-T, the following actions were proposed:

• 
  Recommending establishment of JCT based on N11112, and nomination of chair from ISO/IEC
  
• 
  Recommending to issue CfP jointly
  
• 
  Recommending to start the joint activity by holding a joint meeting in Dresden during 2010/04/15 – 2010/04/23 in order to evaluate the responses to the CfP and initiate steps towards selection of technology
  

An AHG was also appointed (N11120, as recorded in N11076) to further the work until the 92nd meeting.

Input Documents reviewed:

m17162	Comment on Call for Proposals on HVC Requested that: After Pre-registration Information on payment recipient will be conveyed to pre-registered organization if requested. After Formal registration Invoice will be issued after confirming; no possibility of refunding. Payment of the fee: Registered organization shall inform CfP contact person if any delay of payment completion would happen due to procedural issue. These issues were requested to be clarified in the final CfP	National Body of Japan
m17183	UKNB Comments on Call for Proposals on HVC Request to - give more clear statement about the importance of UHDTV, to avoid misinterpretation due to the fact that 4Kx2K is not included in subjective assessment This issue should be worded better in the final CfP - Use decoding time only as rough indicator of complexity Comment noted, but ranking of proposals is to be done by group consensus only after the results of CfP have been evaluated These issues were requested to be clarified in the final CfP - Proceed with VCEG collaboration, stick to the timeline	UK National Body
m17173	7680 x 4320 format test sequences for HVC Offer 8Kx4K 60 fps sequences, 4:2:0 format. 5s each. Locomotive train, Nebuta festival Native resolution of chip is quincunx (red, blue and 2 green channels). Full-resolution 4:4:4 not available currently. Signing of NDA was requested by the contributor – with the contributor organizing distribution, MPEG cannot take responsibility on this (In discussions at the meeting, the NDA request seemed problematic for potential use of the material in MPEG activities.) Longer sequences (10 s) as appropriate for subjective testing could be provided	Shinichi Sakaida Atsuro Ichigaya Yasutaka Matsuo Takahiro Toyoda Kazuhisa Iguchi Yoshiaki Shishikui
m17240	AVC Anchor Streams of Class-A for CfP on HVC Comment: Excel sheets with rates and per-frame PSNR values should be provided for all anchors (before end of the week)	Kazuo Sugimoto Shun-ichi Sekiguchi Keiichi Chono Atsuro Ichigaya Kazuhisa Iguchi Takahiro Nishi Steffen Wittmann
m17162	Comment on Call for Proposals on HVC Request that: After Pre-registration Information on payment recipient will be conveyed to pre-registered organization if requested. After Formal registration Invoice will be issued after confirming no possibility of refunding. Payment of the fee: Registered organization shall inform CfP contact person if any delay of payment completion would happen due to procedural issue. These issues must be clarified in the final CfP	National Body of Japan
m17183	UKNB Comments on Call for Proposals on HVC Request to - give more clear statement about the importance of UHDTV, to avoid misinterpretation due to the fact that 4Kx2K is not included in subjective assessment This issue should be worded better in the final CfP - Use decoding time only as rough indicator of complexity Comment noted, but ranking of proposals is to be done by group consensus only after the results of CfP have been evaluated These issues must be clarified in the final CfP - Proceed with VCEG collaboration, stick to the timeline	UK National Body
m17173	7680 x 4320 format test sequences for HVC Offer 8Kx4K 60 fps sequences, 4:2:0 format. 5s each. Locomotive train, Nebuta festival Native resolution of chip is quincunx (red, blue and 2 green channels). Full-resolution 4:4:4 not available currently. Signing of NDA is requested – Contributors will organize distribution, MPEG cannot take responsibility on this Longer sequences (10 s) as appropriate for subjective testing could be provided	Shinichi Sakaida Atsuro Ichigaya Yasutaka Matsuo Takahiro Toyoda Kazuhisa Iguchi Yoshiaki Shishikui
m17240	AVC Anchor Streams of Class-A for CfP on HVC Comment: Excel sheets with rates and per-frame PSNR values should be provided for all anchors (before end of the week)	Kazuo Sugimoto Shun-ichi Sekiguchi Keiichi Chono Atsuro Ichigaya Kazuhisa Iguchi Takahiro Nishi Steffen Wittmann
m17163	Motion Compensated Prediction by Subsampled Block Matching for Zoom Motion Contents Traditionally only MC for translation motion; other types typically much more complex. Use pre-computed zoom-in and zoom-out frames as additional references. Use the sub-pixel interpolated frames with various re-sampling ratios (typically those which are close to 1, e.g. 8/9 or 9/8 if up-sampling by factor of 8 is done). Turned off all KTA coding tools (using 1/8 pixel MV resolution). BR saving average around 5% for 1/8 pel, 2% for 1/4 pel MC. Additional reference frames are used for signaling.	Lai-Man Po Ka-Man Wong Ka-Ho Ng Kwok-Wai Cheung Chi-Wang Ting Yu Liu Ka-Man Cheng Yan Huo
m17180	Advanced Direct Mode for Bi-directionally Predictive Block Coding Selection of reference blocks for spatial direct is modified depending on block position. Then Motion of a spatial adjacent (as in spatial direct) is used to determine the co-located block, taking into account the motion shift. MV Difference in reference frame also taken into account. Done in both forward and backward directions. Likelihood value is determined for each of the 3 cases, and the case with highest is selected. Advanced direct mode is signaled by additional flag. Old direct mode should not be used. Then, overall bit rate savings compared to alpha anchor is 2.7%; highest for class B (around 3.5). Gain higher for low bit rates.	Siu Wan-Chi (PolyU) Kam Yan-Ho (PolyU) Liu Yu (ASTRI) Cheng Ka Man (ASTRI) Huo Yan (ASTRI)
m17181	Variable Block Size Skip Mode and New Partition Modes for Inter-prediction New MB modes that also allow combination of two 8x8 blocks and one 8x16 block. Skip/nonskip mode also separately for 8x16 and 8x8 block types. Competition based MV selection method (spatial and temporal predictor) investigated, but not used. Comparison against gamma anchor. Average BR reduction 2.4%, more at low bit rates.	Kam Yan-Ho (PolyU) Siu Wan-Chi (PolyU Chan Yui-Lam (PolyU) Hui Wai-Lam (PolyU) Fu Changhong (PolyU) Liu Yu (ASTRI) Cheng Ka Man (ASTRI) Huo Yan (ASTRI)

Output documents: