International organisation for standardisation

6.18Parsing throughput, robustness and error resilience

6.18.1.1.1.1.1.1.1JCTVC-F068 A study on HEVC parsing throughput issue [M. Zhou, V. Sze, Y. Matsuba (TI)]

Considered in BoG coordinated by B. Bross.

6.18.1.1.1.1.1.1.2JCTVC-F699 Cross-check report of TI's proposal JCTVC-F068 on HEVC parsing throughput issue (combined study with JCTVC-F470 part) [T. Sugio, T. Nishi (Panasonic)] [late reg. 07-11, upload 07-11]
6.18.1.1.1.1.1.1.3JCTVC-F074 Cross-check of TI’s study on HEVC parsing throughput issue JCTVC-F068 [B. Bross (Fraunhofer HHI)] [late upload 07-04]
6.18.1.1.1.1.1.1.4JCTVC-F347 A method of decoupling motion data reconstruction from entropy decoding [M. Zhou, V. Sze, Y. Matsuba (TI), T. Sugio, T. Nishi (Panasonic), J. Chen, T. Lee (Samsung), W. Wan, Y. Yu (Broadcom)]

Discussed in BoG coordinated by B. Bross.

6.18.1.1.1.1.1.1.5JCTVC-F470 Parsing Robustness for Merge/AMVP [T. Sugio, T. Nishi (Panasonic)]
6.18.1.1.1.1.1.1.6JCTVC-F090 Cross-verification of Panasonic's proposal JCTVC-F470 on parsing robustness for merge/AMVP [M. Zhou (TI)]
6.18.1.1.1.1.1.1.7JCTVC-F402 MVP index parsing with fixed number of candidates [J. Chen, T. Lee (Samsung)]

Discussed in BoG coordinated by B. Bross.

6.18.1.1.1.1.1.1.8JCTVC-F504 Cross check of Samsung proposal on MVP index parsing with fixed number of candidates (JCTVC-F402) [Guillaume Laroche (Canon), Tangi Poirier] [late upload 07-06]

Discussed in BoG coordinated by B. Bross.

6.18.1.1.1.1.1.1.9JCTVC-F695 Cross-verification of Samsung's proposal JCTVC-F402 on MVP index parsing with fixed number of candidates (combined study with JCTVC-F470 part) [M. Zhou (TI)] [late reg. 07-09, upload 07-10]
6.18.1.1.1.1.1.1.10JCTVC-F550 Removal of the parsing dependency of residual coding on intra mode [J. Sole, Y. Zheng, W.-J. Chien, R. Joshi, X. Wang, M. Karczewicz (Qualcomm)]

This contribution proposed to modify the residual coding in high efficiency usage in order to remove the dependency of the residual data parsing on the intra mode. Three modifications were introduced to attain this goal. First, coefficient level coding is made dependent on the scan position. For this part, the BD-rate for AI-HE, RA-HE and LB-HE configurations is reported as −0.05%, 0.07%, and −0.01%, respectively. Second, the last significant coefficient coding assumes that the scan is always zigzag. Adding this part, the BD-rate for AI-HE, RA-HE and LB-HE configurations is reported as −0.03%, 0.06%, and 0.01%, respectively. Finally, the significance map context for 4×4 and 8×8 blocks are modified to depend on the scan position, not the block-based position as in HM3.0. The BD-rate impact of the three parts combined for AI-HE, RA-HE, and LB-HE is reported as 0.13%, 0.18%, and 0.05%, respectively.

(Discussed in Track P – Thursday 07-21.)

This seems desirable, but may have interactions with other adoptions, so it was deferred for further study. We hope to do something like this.

6.18.1.1.1.1.1.1.11JCTVC-F728 A Cross-check report for JCTVC-F550 proposal removing parsing dependency of residual coding on intra mode Kiran Misra, Andrew Segall (Sharp) [late reg. 07-12, upload 07-12]
6.18.1.1.1.1.1.1.12JCTVC-F302 Merge Candidate Selection in 2NxN, Nx2N, and NxN Mode [Y. Zheng, X. Wang, W.-J. Chien, M. Karczewicz (Qualcomm)]

This contribution proposed a change to the rules currently used in determining merge candidates for a PU under 2NxN, Nx2N or NxN mode. Based on the proposal, when a current PU under these partition modes is not the first PU in a CU, the previous PU’s merge index is checked to determine if motion information of the previous PU can be used as a merge candidate for the current PU. It was asserted that the proposed changes can help solve the parsing robustness issue while preserving coding efficiency.

Compression improvement 0.1%. This contribution was noted to need to be discussed in the BoG on MV coding in the parsing category. The notes were thus moved to that section of the report. May benefit from investigation in CE9.