Back to Search
Document details
Non-EE2: CCP merge mode with adjustment
Abstract
This contribution presents a cross-component prediction (CCP) merge mode with adjustment. In the proposed method, a CCP model can be updated using an adjustment parameter, which is derived based on a template of the current block. On top of ECM-15.0, the simulation results are reported as below:
JVET-AK0223 Non-EE2: CCP merge mode with adjustment [Y. Wang, K. Zhang, W. Yin, Z. Deng, N. Zhang, L. Zhao, M. Salehifar, L. Zhang (Bytedance)]
This contribution presents a cross-component prediction (CCP) merge mode with adjustment. In the proposed method, a CCP model can be updated using an adjustment parameter, which is derived based on a template of the current block. On top of ECM-15.0, the simulation results are reported as below:
AI: {-0.01%, -0.40%, -0.27%; 101.1%, 100.4%};
RA*: {0.00%, -0.29%, -0.16%; 105.7%, 100.3%}.
* Two test points are copied from the anchor and running time for RA is not reliable.
Two additional RDO checks are needed.
Several experts found the method interesting and supported investigation in an EE.
Current tradeoff in terms of encoding time not acceptable – try to improve, where also combination with JVET-AK0202 could give benefit.
Decisions
Current tradeoff in terms of encoding time not acceptable – try to improve, where also combination with JVET-AK0202 could give benefit.
Citation