The video subgroup and the Sunday AHG on AVC development discussed the issues related to the ongoing development of AVC amendments and the corrigendum work related to fifth edition. The most relevant work items in this context were
Study of FPDAM1 (Constrained Baseline, SEI): Some minor changes related to the frame packing arrangement SEI message: Clarification of relationship with aspect ratio, introduction of a row interleaving method
Stereo High Profile: This new profile, which restricts MVC to 2 views but includes interlaced, is strongly supported by 5 companies. Software implementation was done in both JM and MVC reference software, and consistency was shown by cross-checks. In the case off usage for stereo progressive, it is a subset of multiview high profile (using same profile_idc, and introducing constrained_set5_flag for signaling usage of interlaced). The Stereo High Profile is also included in the Study of FPDAM1, and NBs are asked to comment on this. This includes the option of not promoting it into FDAM by July, and start a separate amendment on Stereo High. The video subgroup however recommends to issue this new profile as early as possible, such that possibly a new edition of AVC could be issued after the finalization of Amd.1 and Cor.1.
New version of defect report (planning DCOR1 by July) was also issued. This now also includes a number of corrigendum items related to non-MVC parts of edition 5. Some of the issues that were resolved as a result of the discussion are also explained in the VCEG liaison N10559, but more effort and discussion with VCEG will be necessary until the DCOR can be issued.
Study documents were issued for the FPDAMs on MVC software and conformance (both intended tom reach FDAM status by July)
Integration of AVC related code bases was further discussed. From the implementation of the new Stereo High Profile in both JM and JMVC, it appears realistic that the two code bases could be unified. It must however carefully be checked whether the memory management of JM is sufficient for the purpose of a full MVC (Multiview high with larger number of views) implementation. Unification of JM and JSVM would also be desirable by the longer term, but so far, no effort was made on exploring this possibility.
Documents reviewed:
m16320
On 2D + Depth SEI Message
Follows JVT-AD017 (JVT Geneva Jan 09), now joint proposal of Thomson and Philips.
Supports side-by-side of video-plus-depth (both at same resolution), note: It is certainly not necessary here to allow both options, i.e. it can always be assumed that the image is left and depth is right. stereo-plus depth (left/right images and depth maps at top/bottom of 4-split frame), and LDV (same as before, but foreground/background and their depth maps at top/bottom.
Would be useful to combine it with horizontal squeezing as in the current SEI message from FPDAM1.This could be signaled by sample aspect ratio. Clarification of sample aspect ratio needed there (e.g. horizontal squeezing side by side 2:1, vertical squeezing 1:2), also for the current SEI message in FPDAM1.
No support to the proposal than by the proponents.
Dong Tian
Po-Lin Lai
Fons Bruls
Lincoln Lobo
Wiebe de Haan
m16329
Reference Software And Test Results For Multiview Field High Profile
Implementation for stereo interlaced MVC in both JM and MVC reference software. Results brought for three sequences (Trapeze, Tunnel SD, BMX HD). BR reduction between 3 and 10 percent compared to simulcast, between 5 and 25 compared to MVC frame-based coding (field based coding was not tested).
Software needs more improvement, proponents would do that.
Potentially this could still be included in the FDAM of MVC software. Still to be decided whether the track of JM or MVC ref software should be followed. (recommendation)
ChongSoon Lim
Steffen Wittmann
Takahiro Nishi
m16330
Contribution of Stereoscopic Test sequence
BMX sequence: Copyright conditions provided.
ChongSoon Lim
Steffen Wittmann
Takahiro Nishi
m16331
Standardization of a new MVC profile as specified in Working Draft 1 of ISO/IEC 14496-10:200X/Amd.2 Multiview Field High Profile (N10344)
Support by 5 companies. No change relative to the previous WD. Agreement that this shall be done. More discussion needed whether it could still be included in Amd.1 (then FDAM either July or October) as a Study of FPDAM, or whether this should become a separate amendment (then PDAM now or July, FPDAM October, FDAM 10/04)
Verification of Test Results For Multiview Field High Profile (m16329)
Confirm that results of M16329 match.
Sehoon Yea
Anthony Vetro
Shun-ichi Sekiguchi
m16481
JM Reference Software Enhancements
Adopt this, i.e. Karsten Sühring will publish as JM16 after appropriate checking.
Alexis Michael Tourapis
Athanasios Leontaris
Peshala Pahalawatta
Yan Ye
m16492
Liaison statement from ITU-T SG 16
Technical: Various issues related to AVC corrigendum. Issue of Stereo High Profile: How to make this (in progressive mode) compatible with Multiview High
