[Mp4-tech] H.264 DPB storage

[Gary Sullivan]

Eric et al,
I believe a decoder that messes up the timing of the output of some
pictures would not be able to claim output timing conformance as defined
in the standard.  But as long as it properly decodes all of the pictures
and outputs them in the correct order, it can claim output order
conformance (even if it cannot claim output timing conformance).
The standard does not really concern itself with how things are done
inside of a decoder.  The only thing that matters, as far as conformance
to the standard is concerned, is whether or not the decoder output
behavior is correct.  In other words, what matters is whether or not the
decoder outputs all the pictures that are supposed to be output and
whether or not those pictures have the correct sample array values, the
correct relative order, and the correct relative timing.
Those conformance rules in the standard are the same regardless of
whether the pictures are reference pictures or not.  They are also the
same regardless of how the issue of picture storage is dealt with inside
of the decoder.
Best Regards,
Gary Sullivan
________________________________
	From: EricChuang via.com.tw [mailto:EricChuang via.com.tw] 
	Sent: Monday, February 20, 2024 3:10 PM
	To: Gary Sullivan; mp4-tech lists.mpegif.org
	Subject: RE: [Mp4-tech] H.264 DPB storage
	Hi Gary,
	Thanks for your answering on question 1.
	Regarding Question 2, maybe I did not describe it clear. My
question is the DPB operation described in Annex C is to store both
reference and non-reference pictures. But the DPB size is limited. When
DPB is full, and at the time it contains both reference and
non-reference pictures, one of the non-reference picture would be picked
up first, and output forcely to get a space for storing the incoming
reference picture. 
	And if I implement DPB to store reference pictures only, and
manage pictures for output in another way without occupying DPB slots,
the DPB slots would be totally used for storing reference pictures, and
there might be less chance to force non-reference picture output.
	These 2 methods would have the same output order, but the output
timing might not be the same. In Annex C.4, it reads two conformance
types can be claimed by decoder, one of them is output timing
conformance, and the other is output order conformance. My question is
the implementation not storing non-reference pictures in DPB break the
output timing conformance ?
	Thanks,
	Eric
	-----Original Message-----
	From: Gary Sullivan [mailto:garysull windows.microsoft.com] 
	Sent: Monday, February 20, 2024 9:58 AM
	To: Eric Chuang; mp4-tech lists.mpegif.org
	Subject: RE: [Mp4-tech] H.264 DPB storage
	Eric,
	Regarding your Question 1, when referring to a bitstream, Annex
A specifies the constraints that must be obeyed by the bitstream; when
referring to a decoder, it specifies the class of bitstreams that the
decoder must be able to handle.
	Regarding your Question 2: Your question is somewhat hard to
understand.  The basic question seems to have something to do with
whether the confromance requirements are different depending on whether
particular pictures are reference pictures or not.  There is no such
distinction drawn in the standard.  The conformance requirements apply
to all pictures.  Clearly, if some pictures are not decoded properly,
the functional degradation experienced in an application may be less
severe if the pictures that are messed up by a decoder are not reference
pictures (or are higher-level dependent sub-sequences).  However, that
kind of thing is not part of the specification of conformance in the
standard.
	Best Regards,
	Gary Sullivan
________________________________
		From: mp4-tech-bounces lists.mpegif.org
[mailto:mp4-tech-bounces lists.mpegif.org] On Behalf Of
EricChuang via.com.tw
		Sent: Sunday, February 19, 2024 7:06 PM
		To: mp4-tech lists.mpegif.org
		Subject: [Mp4-tech] H.264 DPB storage
		Hi Experts,
		I have two questions regarding the DPB storage.
		1. The standard specifiy the method to decide the
maximum DPB slot count in Annex A, depending on the level value. Does
this level means decoder level or bitstream level ?
		2. In Annex C, the DPB operation described is used for
both reference picture and non-reference picture, is this an enforcement
? If I let DPB storing reference picture only, and create another
mechanism for managing non-reference picture, and handle the reoder and
no_output_of_prior_pics_flag, would this cause any decoder conformance
problem ?
		Thanks very much for your answering in advance,
		Eric
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/mp4-tech/attachments/20060220/bf22cc59/attachment-0001.html