JVET-O0288 CE5-related: On the syntax constraints of ALF APS [O. Chubach, C.-Y. Chen, T.-D. Chuang, C.-W. Hsu, Y.-W. Huang, S.-M. Lei (MediaTek)]

This contribution proposes four cleanups for the adaptive loop filter (ALF) syntax in slice header and adaptive parameter set (APS). The first cleanup is removing the constraints for a number of APSs referenced in one slices. In the current VVC, the maximum number of APSs referenced in one slice depends on slice type, and this contribution proposes to use the same maximum value for all slice types. The second cleanup is removing the inference of APS ID referenced in one slice for chroma (slice_alf_aps_id_chroma) when only one APS is referenced for luma ALF and the slice type is I slice. The third cleanup is adding one encoder normative constraint that the APS referenced in one slice for luma/chroma shall contain the luma/chroma filter parameters at least. The last cleanup is inferring the flag specifying whether a chroma filter is signalled (alf_chroma_filter_signal_flag) to be true, when luma filter parameters are not signalled (alf_luma_filter_signal_flag=0) in the same APS, in order to avoid signalling one ALF-type APS without any ALF parameters. The proposed modifications are implemented on top of VTM5.0. It is reported that the four cleanups together achieves 0.00%/0.00%/0.00%, 0.00%/0.00%/0.00%, and 0.00%/0.00%/0.00% Y/Cb/Cr BD-rates for AI, RA, and LB, respectively.

The first aspect is unifying the max number of ALF APS for inter and intra (currently, it is 1 for intra and 6 for inter). Agreed.

It is discussed that a practical max number is still to be decided.

The second aspect is to remove the inference of chroma APS ID from luma. Following the discussion under JVET-O0247, this is agreed to be useful. Agreed. Decision: Adopt JVET-O0288 aspects 1 and 2

The third aspect suggests that if an ALF APS is referenced in the slice header, it should at least contain filter coefficients for at least one filter (encoder / bitstream constraint to be formulated both for luma and chroma). It is noted that the spec already specifies that in this case the filter coefficients are inferred as zero. No action necessary.

The fourth aspect suggests that in an APS at least either luma or chroma shall be non-zero. There is however no problem with the current design, it is up to the encoder not to do that, decoder behaviour is still well defined.

(The following statement does not apply to this specific contribution)

It is noted that, as a side activity to update the ALF APS, plus correspnding slice level and CTU level aspects, the different proponents have circulated their text pieces and coordinated with B. Bross.