JVET-AD0136 AHG10: Lagrange multiplier optimization for chroma ALF and CCALF [W.-J. Zou, Y. Zhou, C.-M. Gu, J.-W. Fan (Xidian Univ.), S.-W. Xie, G. Ying, M.-H. Jia, Y.-L. H, C. Huang (ZTE)]

This contribution presents an RDO optimization for ALF. The chroma Lagrange multiplier is adjusted by a scaling factor when the luma ALF APS transmission flag is ON and the number of history chroma ALF APS is insufficient. It is reported that, with VTM19.0 as anchor and test platform, the overall performance was as follows:

• LDB: 0.02%, -0.87%, -1.19%

Class B: 0.07%,-0.91%,-1.43% Class C: 0.03%,-0.31%,-0.67%

Class E: -0.09%,-1.55%,-1.48%

The cross checker reported that the proposed changes will cause a drop in performance for class F, and commented that the proposed change should not be used on screen content.

The proposed method will affect our CTC performance.

The proposed method is asserted to bring more benefit to the LDB configuration because the APS buffer is never reset, and the identified problem would occur more. The proponent suggests to only consider this method for LDB.

The cross checker reported that encoding and decoding runtime are not affected (both 100%).

It was commented that LDP would also be in a similar situation as LDB, but no results for LDP are available. It was commented by VTM software coordinator that it is more desirable to keep LDB and LDP aligned with regard to this proposed method.

It was commented that the proposed method causes some performance loss on top of ECM. It was suggested to investigate this and improve the way this method operates on ECM.

Further study was recommended.