Back to Search Document details
16th Meeting: Geneva, October 2019 2019-09-17 13:06
CE6-2.1: LFNST with 3 transform sets
Abstract
This contribution reports the coding performance of CE6-2.1, which simplifies LFNST by reducing four transform sets to three transform sets (i.e., removing transform set 2) and saves 25% memory for storing LFNST matrices, i.e., 2KB. It is reported that the change results in 0.03% / 0.06% / 0.11% BD-rates in AI configuration for Y / Cb / Cr, and -0.01% / 0.06% / 0.03% BD-rates in RA configuration for Y / Cb / Cr.
JVET-P0044 CE6-2.1: LFNST with 3 transform sets [Y. Zhao, H. Yang (Huawei)]

This contribution reports the coding performance of CE6-2.1, which simplifies LFNST by reducing four transform sets to three transform sets (i.e., removing transform set 2) and saves 25% memory for storing LFNST matrices, i.e., 2KB. It is reported that the change results in 0.03% / 0.06% / 0.11% BD-rates in AI configuration for Y / Cb / Cr, and -0.01% / 0.06% / 0.03% BD-rates in RA configuration for Y / Cb / Cr.

Decisions
This contribution reports the coding performance of CE6-2.1, which simplifies LFNST by reducing four transform sets to three transform sets (i.e., removing transform set 2) and saves 25% memory for storing LFNST matrices, i.e., 2KB. It is reported that the change results in 0.03% / 0.06% / 0.11% BD-rates in AI configuration for Y / Cb / Cr, and -0.01% / 0.06% / 0.03% BD-rates in RA configuration for Y / Cb / Cr.
Citation