Back to Search Document details
29th Meeting: by teleconference, DE, January 2023 2023-01-13 17:42
EE2-related: Modifications of MTS and LFNST for IntraTMP
Abstract
In ECM-7.0, an IntraTMP coded block uses block size dependent implicit MTS kernels as primary transform and LFNST transform set 0 as the secondary transform. Intra TMP coded blocks are assigned planar mode as the intra prediction mode. In JVET- AB0115 (EE2-4.1), it is proposed to derive the intra prediction mode for IntraTMP block using DIMD method.
JVET-AC0205 EE2-related: Modifications of MTS and LFNST for IntraTMP [R. G. Youvalari, D. Bugdayci Sansli, J. Lainema (Nokia)] [late]

In ECM-7.0, an IntraTMP coded block uses block size dependent implicit MTS kernels as primary transform and LFNST transform set 0 as the secondary transform. Intra TMP coded blocks are assigned planar mode as the intra prediction mode. In JVET- AB0115 (EE2-4.1), it is proposed to derive the intra prediction mode for IntraTMP block using DIMD method.

This contribution proposes to inherit the intra prediction mode of an IntraTMP block from its reference block or from the neighbourhood of its reference block. The proposed method has a simpler approach for selecting the intra prediction mode and it is reported to provide slightly higher coding gain than the EE tests.

The impacts on coding efficiency and runtimes over ECM-7.0 are reportedly {for Y, U, V, EncT, DecT}:

Test A:

AI { -0.05%, -0.01%, -0.07%, 101%, 101%}, RA { -0.02%, -0.08%, -0.06%, 98%, 95%}

Test B (in addition to Test A, the derived intra prediction mode is used in MPM list of neighbouring blocks):

AI { -0.06%, -0.03%, -0.01%, 101%, 101%}, RA { -0.xx%, -0.xx%, -0.xx%, 1xx%, 1xx%}

There was no obvious advantage to this over the adopted EE2 4.1a, so no action was taken on it.

Decisions
adopted
There was no obvious advantage to this over the adopted EE2 4.1a, so no action was taken on it.
Citation