[Mp4-tech] Is it possible to separate chroma prediction and lumapredictioninintra prediction of H.264 ?

[Gary Sullivan]

Richard Park et al,
In the standard, chroma intra prediction is controlled by one or more
separate syntax element(s) from what controls luma intra prediction.
However, there can be other secondary interactions between other parts
of luma and chroma intra coding.  For example, I think
coded_block_pattern is affected by both luma and chroma.  Also, there is
strong interaction between the residual coding of luma and chroma when
the residual color transform (RCT) is in use.
I think the problem of slow encoding due to exhaustive joint exploration
of luma and chroma intra prediction modes may have been fixed in a more
recent version of the software than the status that JVT-N014 reflects.
I think the cross-dependency was introduced for the benefit of RCT
encoding performance, but RCT does not apply in most scenarios.  So I
think the software was later altered to only do exhaustive interaction
testing between luma and chroma intra predictions when RCT is being
used.
Best Regards,
Gary Sullivan
________________________________
	From: mp4-tech-bounces lists.mpegif.org
[mailto:mp4-tech-bounces lists.mpegif.org] On Behalf Of Park Jun Sung
	Sent: Tuesday, August 30, 2023 4:42 PM
	To: mp4-tech lists.mpegif.org
	Subject: [Mp4-tech] Is it possible to separate chroma prediction
and lumapredictioninintra prediction of H.264 ?
	Hi, experts.
	I have a question about Chroma intra prediction in H.264.
	The current structure of the codec forces all intra decisions to
be performed for each possible Chroma intra prediction mode. 
	As a result, the computation complexity is significantly
increased.
	So the flag, ChromaIntraDiable, is introduced in JVT-N014 (Jan.,
2005) to reduce complexity by allowing DC Chroma prediction only.
	It is a kind of temporal technique because there is an intention
of fixing the JM software but lack of time.
	(JM is focusing to optimize rate/distortion, so it is needed to
enhance the complexity.)
	Now the question,
	Is it possible to totally separate Chroma prediction from Luma
prediction in intra prediction?
	If each of Chroma prediction and Luma prediction performs
independently, the computation complexity will be decreased greatly.
	Thanks in advance.
	Regards,
	Richard Park
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/mp4-tech/attachments/20050831/fcf8ec7b/attachment.html