Back to Search Document details
30th Meeting: Antalya, TR, April 2023 2023-04-21 08:52
Non-EE2: Extension of IBC-GPM
Authors: Changyue Ma Kwai Xiaoyu Xiu Kwai Wei Chen Kwai Hong-Jheng Jhu Kwai Che-Wei Kuo Kwai Ning Yan Kwai X. Wang Kwai
Abstract
In ECM-8.0, IBC-GPM is applied which divides a CU into two geometry partitions based on a sample-wise combination of two prediction signals. In the current design, one of the IBC-GPM prediction signals is generated from the IBC prediction while the other is generated from the intra prediction. In this contribution, the existing IBC-GPM is further extended to allow the predictions of both partitions are generated using the IBC. The simulation results on top of ECM-8.0 are summarized as follows.
Meeting Notes
JVET-AD0215 Non-EE2: Extension of IBC-GPM [C. Ma, X. Xiu, W. Chen, H.-J. Jhu, C.-W. Kuo, N. Yan, X. Wang (Kwai)]

In ECM-8.0, IBC-GPM is applied which divides a CU into two geometry partitions based on a sample-wise combination of two prediction signals. In the current design, one of the IBC-GPM prediction signals is generated from the IBC prediction while the other is generated from the intra prediction. In this contribution, the existing IBC-GPM is further extended to allow the predictions of both partitions are generated using the IBC. The simulation results on top of ECM-8.0 are summarized as follows.

AI:

Class F : {-0.16%, -0.27%, -0.28%, 100%, 100%}, Class TGM : {-0.43%, -0.34%, -0.31%, 99%, 99%};

RA:

Class F : {-0.17%, -0.86%, -0.25%, 100%, 101%}, Class TGM : {-0.36%, -0.18%, -0.25%,100%, 100%};

It was agreed to investigate this in an EE (along with JVET-AD0134 which proposes a similar approach)

Cited:
JVET-AD0372 X
ARCHIVE :
JVET-AD0215-Table1.xlsm JVET-AD0215.docx JVET-AD0215.pptx JVET-AD0215_r1.docx
Decisions
It was agreed to investigate this in an EE (along with JVET-AD0134 which proposes a similar approach)
Citation