This contribution proposes several simplification methods for OBMC. The first method is to remove the ‘obmc_flag’ syntax which is signalled in every CU. Encoder would save memory associated with OBMC on/off decision. The second method is to disable a sub-block OBMC according to the result of a template matching based OBMC. If a template matching based OBMC has not been performed on any CU boundary block, a sub-block OBMC is also not applied for the CU. The third method is to disable OBMC in coding blocks with large block size. When the block size cbW * cbH is greater than or equal to 32(ECM-5.0) and cbW and cbH is less than 128 (additional restriction), OBMC is applied. On top of ECM-5.0, simulation results of the proposed method are reported as below:

This document reports the crosschecking results of JVET-AA0143 on simplification methods for OBMC. Specifically, three modifications are proposed for the existing OBMC design in the ECM5.0: 1) removing the ‘obmc_flag’ syntax at CU level; 2) disabling a sub-block OBMC based on template matching OBMC; 3) disabling OBMC for large block sizes.

This contribution reports the crosscheck results of JVET-AA0073. The partial simulation results confirm that BD-rate results match with the proponent’s results under allintra and randomaccess test conditions. The slight runtime differences are observed in few cases which may be due to the variation of CPU clusters.