JVET-Q0472 AHG15: Quantization matrix signalling [P. de Lagrange, F. Le Léannec, E. François, K. Naser (InterDigital)]
This contribution proposes two changes to quantization matrix signalling.
In a first aspect, it removes the scaling_list_pred_mode_flag syntax element, which is said to be mostly redundant.
In a second aspect, scaling_list_copy_mode_flag is replaced by a number (scaling_list_num_coef), which specifies the number of delta-coefficient transmitted, possibly less than matrix size, the missing ones being inferred to zero. This is said to enable efficient QM adjustments at low bit cost.
The second aspect had already been proposed by the last meeting and not been adopted. There is no new evidence that change of quant matrices per picture would be useful, which would make such large compression necessary.
Nothing is broken in the signalling of the quantization matrix. No urgent need for action. Stability of the specification has higher importance.