Back to Search
Document details
Non-CE7: A cleanup for inter_pred_idc coding
Abstract
It is asserted that the current design of inter_pred_idc coding is inconsistent, because the same context model is shared by the coding of the bi_pred_flag for 8x8 CUs and the list0/list1 flag of uni-pred CUs of all sizes, two basically unrelated elements. It is proposed to add one context model so that the different context models can be assigned to coding of those two elements. The experimental results revealed that the proposed change led to an average BD-rate difference of -0.01% in RA and -0.02% in LD_B, respectively, when compared to the VTM6.0 anchor.
JVET-P0042 Non-CE7: A cleanup for inter_pred_idc coding [B. Heng, M. Zhou (Broadcom)]
It is asserted that the current design of inter_pred_idc coding is inconsistent, because the same context model is shared by the coding of the bi_pred_flag for 8x8 CUs and the list0/list1 flag of uni-pred CUs of all sizes, two basically unrelated elements. It is proposed to add one context model so that the different context models can be assigned to coding of those two elements. The experimental results revealed that the proposed change led to an average BD-rate difference of -0.01% in RA and -0.02% in LD_B, respectively, when compared to the VTM6.0 anchor.
Though this rather relates CE4 (MV coding), it was presented in CE7.
Decision (cleanup): Adopt JVET-P0042.
JVET-P0541 touches a similar issue.
References:
Decisions
JVET-P0541 touches a similar issue.
Citation