3Project development, status, and guidance (34)
3.1Communication to and by parent bodies (0)
qq
3.2Conformance test set development (2)
JCTVC-O0084 Editor's proposed draft text of HEVC conformance testing [T. Suzuki, G. Sullivan, W. Wan]
Reviewed Thu 31st a.m. (GJS).
Text was proposed by the editors to update the available bitstream information.
JCTVC-O0104 Corner case conformance streams [C. Fogg (Harmonic)]
Reviewed Thu 31st am (GJS).
For the unfulfilled conformance bitstreams, it was proposed that JCT volunteers collaborate to create a modified HM encoder to generate them. This proposal outlines suggested methods to most quickly achieve this goal, including use of stream element profiling, and iterative frame synthesis to trigger the desired corner cases.
The contribution suggested to use dynamic lists to form counters for each element combination.
If a test pattern is not found in a sweep of the existing conformance suite, the contribution suggested that JCT volunteers could create such bitstreams, in a collaborate project based off the HM encoder, if the case is considered important.
The presenter was not available for review.
In the discussion, it was suggested that there could be some configuration parameters to use within the HM codebase that would trigger the generation of corner cases by the HM. It was suggested that if we do this, checking whether some change breaks this functionality could become part of the check-in process.
3.3Version 1 bug reports and cleanup (1)
JCTVC-O0054 Editors' proposed corrections to HEVC version 1 [Y.-K. Wang, G. J. Sullivan, B. Bross]
Reviewed Tues. 29th (GJS).
Detailed review was not deemed necessary. Decision (Ed.): Approved as our current state of errata reporting, to be integrated into RExt when it is final. One addition correction was noted and agreed, which was changing "nominal" to "extended" in the semantics of referenceref_screen_luminance_white.
See also notes for O0002 (esp. regarding quantization groups, which should be further studied to determine whether action is needed).
3.4RExt text improvements (3)
JCTVC-O0042 Proposed Editorial Improvements to High efficiency video coding (HEVC) Range Extensions Text Specification Draft 4 [D. Flynn (BlackBerry)] [miss]
Reviewed Fri. 1st (GJS). Detailed review was not deemed necessary. Decision (Ed.): Approved as our current state of RExt basis text development upon which to further integrate for subsequent work.
The contribution discussed an issue regarding down-shifting of reconstructed transform coefficients at high bit depths (ticket #1185). Decision: The #1185 suggested modification (from N0188) was adopted.
JCTVC-O0133 Text fixes for inter RDPCM in HEVC-Rext [M. Naccari, M. Mrak (BBC)]
Discussed in RExt BoG O0352.
JCTVC-O0192 AHG5/AHG6/AHG9: Text changes for HEVC Range Extensions [P. Lai, S. Liu, S. Lei (MediaTek)]
Discussed in RExt BoG O0352.
3.5HEVC coding performance, implementation demonstrations and design analysis (13)
3.5.1Version 1 verification test (1)
JCTVC-O0267 Proposed updates to HEVC verification test plan draft 1 [T.K. Tan (NTT DOCOMO)]
Reviewed Tues. 29th (GJS).
This document contains the draft plan for the video verification test to be conducted to verify the coding performance of the HEVC Main and Main 10 profiles. A formal subjective evaluation will be conducted comparing the HEVC Main and Main 10 profiles to the AVC High and High 10 profiles, respectively. A range of video resolutions from 480p to 4K will be tested.
This document contains proposed updates to the JCTVC-N1011 document. Test sequences and encoding bitrates for use in the verification test plan weare proposed for discussions and consideration at the JCTVC meeting in Geneva.
See also notes on AHG report O0019.
3.5.2RExt coding performance (5)
JCTVC-O0101 HDR coding results [C. Fogg (Harmonic)]
Discussed in RExt BoG O0352.
JCTVC-O0366 Cross-check Report for JCTVC-O0101 on High Dynamic Range Video Coding Results [P. Yin, T. Lu] [late]
JCTVC-O0366 Cross-check Report for JCTVC-O0101 on High Dynamic Range Video Coding Results [XXX]
Discussed in RExt BoG O0352.
JCTVC-O0148 RExt: Testing results on 12-bit 4:2:0 Traffic Sequence [M. Zhou, P. Chen (Broadcom)]
Discussed in RExt BoG O0352.
JCTVC-O0184 Comparison of Compression Performance of HEVC Range Extensions Draft 4 with AVC High 4:4:4 Predictive profile [B. Li, J. Xu, G. J. Sullivan (Microsoft)]
Discussed in RExt BoG O0352.
JCTVC-O0258 AHG8: Results from coding 4:4:4 screen content sequences [H. Yu, Z. Ma, J. Ye, X. Wang, W. Wang (Huawei)] [late]
Discussed in RExt BoG O0352.
(Initial version rejected as placeholder; corrected on 10-16.)
JCTVC-O0269 AhG8: Using qp=0, 4, 8, 12 for visually lossless coding experiments and results from SC coding by packed pixel pseudo-2D-matching integrated with HM12.0RE4.0 [Tao Lin, Xianyi Chen, Peijun Zhang, Liping Zhao (Tongji Univ.)]
Discussed in RExt BoG O0352.
JCTVC-O0337 Caution required when evaluating the performance of screen content sequences [D. Flynn, D. He, G. Martin-Cocher (Blackberry)] [late]
Discussed in RExt BoG O0352.
3.5.3SHVC coding performance (2 - done)
JCTVC-O0097 AHG10/AHG12: On SHVC common test conditions [Y. He, Y. Ye (InterDigital), V. Seregin (Qualcomm)]
Reviewed Tues 29th (GJS).
This contribution proposes to restore previous SHVC common test conditions to simplify the software implementation and be consistent with HM configurations. It was asserted that the proposed change can be simply achieved by disabling the macro in the SHM software.
At the last meeting, there was a (non-normative, encoder-only) change to reference picture list construction (associated with a macro FINAL_RPL_CHANGE_N0082). It was reported that this introduced difficulties in working with the software, and had no significant effect on compression.
Decision (S/W): Disable the encoder-only change macro (FINAL_RPL_CHANGE_N0082).
It was remarked that default reference picture list construction has been proposed to be modified in some way, and the proper impact of this on CTC hads not been determined.
JCTVC-O0243 Cross-check report of AHG10/AHG12: On SHM Common Test Conditions (JCTVC-O0097) [O. Nakagami (Sony)]
Reviewed Tues 29th (GJS).
Confirmed that there is no significant impact for the proposed change.
3.5.4SHVC design aspects (2 - done)
JCTVC-O0115 Pipeline and parallel architecture for the SHVC decoder [W. Hamidouche, M. Raulet, O. Deforges (IETR/INSA)]
Reviewed Tues 29th (GJS).
This contribution describes an implementation of the SHVC decoder in a "GPAC" player context. The SHVC decoder is based on the open source software "OpenHEVC", which reportedly implements a conforming HEVC decoder (2x-5x faster than HM decoder). The proposed pipeline and parallel SHVC decoder enables two levels of parallelism. The first level decodes the base layer and the enhancement layer frames in parallel. The second level of parallelism performs the decoding of both the base layer and the enhancement layer in parallel by taking advantage of the HEVC wavefronts and tiles high-level parallel processing features.
Experimental results carried out on a computer fitted with an Intel Xeon processor running at 3.4 GHz reportedly showed that the pipeline SHVC decoder achieves the decoding of 1920x1080 base layer and 2560x1600 enhancement layer at 45 frames per second when using six concurrent threads.
This type of work was suggested to be complementary to that described in O0165, since this one focused on tiles & wavefronts and the other focused on the slice level and CTB level.
JCTVC-O0165 Real time SHVC software decoding with multi-threaded parallel processing [S. Gudumasu, Y. He, Y. Ye, Y. He (InterDigital)]
Reviewed Tues 29th (GJS).
This contribution describes a parallel decoding framework for SHVC. Various optimization technologies were reportedly implemented on the basis of SHM-2.0 software to achieve real time decoding speed. Multi-threading and SIMD optimizations were reportedly applied to accelerate the decoding process.
Experiments on a desktop with Intel i7 processor 2600 running at 3.4 GHz reportedly show that the parallel SHVC software decoder is able to decode 1080p spatial 2x scalability up to 60 fps (frames per second) and 1080p spatial 1.5x scalability up to 50 fps. The bitstreams were reportedly generated with SHVC common test conditions.
One thread was used for entropy decoding and multiple threads were used for decompression based on "CTB groups" and an "application thread manager" was used for control.
SIMD was used for MC, IQ, IT, and reconstruction.
The contributor suggested that this confirms that the reference index approach provides a suitable design for real-time decoding.
3.5.5Implementation demonstrations (2)
JCTVC-O0321 HEVC interlaced coding – application software proposal [Zineb Agyo, Jérôme Vieron, Jean-Marc Thiesse, Pierre Larbier] [late]
Presenter not available – contribution is available for study.
JCTVC-O0100 Field/frame sequence coding [C. Fogg (Harmonic)]
Presenter not available – contribution is available for study.
Dostları ilə paylaş: |