Re: [Cellar] AV1 mapping Matroska

Hi,

2018-06-30 0:09 GMT+02:00 Andreas Rheinhardt
<andreas.rheinhardt@googlemail.com>:
> Hello,
>
> I have only just started reading the specification, but I already have
> some comments:
>
> 1. Your proposal says that DisplayHeight and DisplayWidth must be set so
> that they agree with what is said in the CodecPrivate/in the bitstream
> so that one can't use the Matroska fields anymore to override what is
> set in the bitstream. If I am not mistaken, then AV1 would be unique in
> this regard among all the video codecs that are allowed in Matroska. I
> don't see a reason for this at all and propose that there shouldn't be a
> special treatment for DisplayHeight and DisplayWidth in AV1.

You're right. This is a mistake on my part. Instead of MUST it's a
SHOULD. Because Matroska allows to override the codec internal values
to display this differently, including changing the aspect ratio or
forcing physical dimensions. I will change that.

> 2. Two sequence header OBUs with the same [seq_profile],
> [high_bitdepth], [seq_level_idx] and [color_range] can still differ in
> the [BitDepth] derived from them due to [twelve_bit]. Therefore I wonder
> if this is the right restriction. Wouldn't it be better if it is
> demanded that the derived [BitDepth] coincide for every sequence header OBU?

It depends if it influences the output of the decoder or the decoder
that can be used. My main concern here is that we have the necessary
information to setup the decoder to start playback and be done with
the duration of the Segment. Otherwise that means the CodecPrivate is
sufficient.

There's a fine line between being too restrictive and forcing multiple
Segment for a simple internal change in the codec. So we don't want to
restrict too many field from the Sequence Header. But on the other
hand we want to ensure a Segment is consistent, especially regarding
seeking.

That being said, you're right [high_bitdepth] and [twelve_bit] are the
ones needed to make up the actual [BitDepth] so I will change that.

> 3. Given that AV1 seems to only use one Sequence Header OBU at a time
> couldn't one use the CodecState and CueCodecState elements to designate
> where the currently active Sequence Header OBU can be found so that one
> doesn't need to repeat the Sequence Header with every keyframe?

Yes, that would be a good use for when seeking is used. But these
elements were ruled out as deprecated because they were of no use
until now. I know VLC doesn't read them for example.

> (Btw: There are currently conflicting recommendations regarding the
> existence of Sequence Header OBU if there is actually only one Sequence
> Header OBU. On the one hand, they should be omitted, on the other hand
> every block marked as keyframe should start with a Sequence Header OBU.)

You mean in my document or the AV1 specs ? My understanding is that
the AV1 bitstream allow Sequence Header OBUs to be found within the
stream. And doesn't mention which elements are allowed to change or
not. Maybe there will be profiles for such restrictions. It won't be
stored in the bitstream (which is frozen) but could be extra data we
put in the CodecPrivate.

> Andreas Rheinhardt/mkver
>
> PS: This is my first mail to this list and I hope that it will be
> presented as a reply to Steve Lhomme's proposed AV1 mapping in Matroska,
> but given the fact that I only subscribed today I can't simply reply,
> but had to explicitly set the "In-Reply-To" field. Hope it works.

It worked perfectly. Thanks for taking the time to review this !

> _______________________________________________
> Cellar mailing list
> Cellar@ietf.org
> https://www.ietf.org/mailman/listinfo/cellar

-- 
Steve Lhomme
Matroska association Chairman