Re: [AVT] [MMUSIC] spatial-resolution parameter for RFC 3984bis

[Kyunghun Jung <kyunghun.jung at samsung.com>]

Title: Samsung Enterprise Portal mySingle

Dear experts of IETF MMUSIC and AVT:

 

Let me add a few comments from 3GPP's point of view.

In fact, we asked MMUSIC to build "Imageattr" for 3GPP's IMS-based video telephony.

 

(1) Codec Dependency

We need a generic, codec-independend mechanism  to negotiate image size between mobile phones.

Parameter sets of recent codecs include the information on image size but they are not considered during session

setup. How to couple imageattr and config (for MPEG-4) or sprop-parameter-set (H.264) will be the key issue

in the session setup. To reduce session setup delay, the offerer might include multiple lines of parameter sets

in SDP offer (one set for for each image size). It might look quite messy.

 

(2) Frame Rate

For mobile video telephony like 3GPP's, frame rate is not negotiated but decided by the sender after the bit-rate

and image size are fixed. Note also that even in wired video telephony, H.245 videoTemporalSpatialTradeoff or "a=quality"

works properly, only between proprietary terminals that know the far-end's capability and codec configuration completely.

 

However, for us, frame rate is also a parameter to be controlled during communication. When there is congestion in downlink

(base station to mobile) or somewhere else, receiving mobile asks the sending mobile to reduce the bit-rate, which can be

executed by reducing the frame rate.

 

(3) For Mr. Even Roni

 1. Announce sending of the specific resolution - just resolution 2. Request a specific resolution - Receiver wants

a specific resolution 3. Negotiate available resolutions - I think the img attribute address 1 and 2, I am not sure about 3

Thank you for the comment. For 3GPP, 3 is the key reason why we need imageattr.

If 3 is still unclear to any MMUSIC/AVT experts, immediate action will be necessary to fix this issue during IETF '72.

Based on the feedback from IETF, 3GPP SA4 will cover imageattr more in depth from August 18.

 We will appreciate if you point out the area you are still not sure yet.

 

Other comments and questions on imageattr will be greatly welcomed and answered.

We would like to ask experts of MMUSIC and AVT to discuss "a=imageattr" with special interest, during IETF '72.

 

Sincerely yours,

Kyunghun Jung

Samsung Electronics Co., Ltd.

------- Original Message -------
Sender : Ingemar Johansson S<ingemar.s.johansson at ericsson.com>
Date   : 2008-07-28 22:01 (GMT+09:00)
Title  : Re: [MMUSIC] [AVT] spatial-resolution parameter for RFC 3984bis

Hi

My response below, also a crosspost to mmusic, hope that this not too upsetting. 

To answer Roni's and Ye-Kui's comments I believe that the image attribute can be used for 3. However, looking in perspective I am not longer sure that the possibility to specify a framerate range makes sense as it does not connect  too wellto reality when it comes to e.g H.264 which talks about macroblocks per second. The conclusion is then that the framerate (or MBs) is determined from the fmtp line.
This means that if the imageattr specifies a large offered range the frame rate is limited by what is possible for the given preferred resolution. 

As an example, consider H.263 with the fmtp line
 CIF=4;QCIF=3;SQCIF=2;
the imageattr may then be specified as
a=imageattr:97 [x=[128:8:352],y=[96:8:288],par=[1.2,1.3]]
If the receiver prefers the resolution 344x288 (slightly less than CIF) the maximum frame rate would be 30/(1.0001*4) ~ 7.5Hz. 
If the prefered resolution is 160x128 (slightly less than QCIF) then the max frame rate is 30/(1.0001*3) Hz and so on.
 
For the H.264 case the frame rate depending of the preferred image size would be limited according to the maximum framerate repending on the max MBs and the resolution given in table A-6 in the H.264 spec. For instance for level 2 an image size of 344x280 which is less than CIF but larger than QVGA would yield a max frame rate of 30Hz.

Would this distincion between framerate (or MBs) and preferred image size work?

/Ingemar   



> -----Original Message-----
> From: Even, Roni [mailto:roni.even at polycom.co.il] 
> Sent: den 28 juli 2008 11:32
> To: Ye-Kui.Wang at nokia.com; Ingemar Johansson S; avt at ietf.org
> Cc: rjesup at wgate.com
> Subject: RE: [AVT] spatial-resolution parameter for RFC 3984bis
> 
> Hi,
> I think that we have to be clear about the usage of the parameter
> 
> 1. Announce sending of the specific resolution - just 
> resolution 2. Request a specific resolution - Receiver wants 
> a specific resolution 3. Negotiate available resolutions - 
> The offer says, I can do CIF, QCIF, 1080P,... here it may 
> dependent in H.263 and H.261 on the frame rate while in H.264 
> the limit is in the macroblocks per second which will specify 
> the maximum frame rate. The reason for this offer is because 
> it offers the operation points that the offerer can have. The 
> level parameter in H.263 and H.264 gives a maximum but does 
> not imply support for any resolution that the answer would 
> like to receive that falls within the range of the level.
> 
> I think the img attribute address 1 and 2, I am not sure about 3
> 
> Roni Even
> 
> > -----Original Message-----
> > From: Ye-Kui.Wang at nokia.com [mailto:Ye-Kui.Wang at nokia.com]
> > Sent: Monday, July 28, 2008 1:11 PM
> > To: ingemar.s.johansson at ericsson.com; avt at ietf.org
> > Cc: Even, Roni; rjesup at wgate.com
> > Subject: RE: [AVT] spatial-resolution parameter for RFC 3984bis
> > 
> > 
> > Interesting! I just wrote the following email, planned to 
> send to the 
> > MMUSIC mailing list. Now this comes only to AVT. Feel free 
> to forward 
> > it to MMUSIC (don't know whether it is allowed to cross posting) or 
> > let me to send that, if needed.
> > 
> > 
> > After a reading of this draft (draft-johansson-mmusic-image- 
> > attributes), I think the sprop-spatial-resolution parameter in the 
> > RFC3984bis can go. It is completely covered by imageattr.
> > 
> > A couple of comments to draft-johansson-mmusic-image-attributes.
> > 
> > - It would be helpful to introduce the methods in payload 
> format specs 
> > for H.261, H.263 and H.264. In particular, for the one I am 
> familiar 
> > with (i.e. H.264), I am not sure how exactly image size 
> together with 
> > frame rate is negotiated.
> > 
> > - Re the following paragraph, I am not sure which parameters are 
> > "similar features" and MUST NOT be specified in imageattr. 
> Basically,
> > H.264 sequence parameter set (SPS)may contain all 
> information relating 
> > to what may be included in imageattr according to the curent draft. 
> > For example, the VUI part of SPS may even contain frame 
> rate related 
> > synatx elements.
> >    o  Conflict with sprop-parameter-set: H.264 defines the sprop-
> >       parameter-set which may define features that are 
> defined in this
> >       draft.  To avoid conflicts, sessions that use the imageattr
> >       attribute MUST NOT specify similar features in 
> sprop-parameter-
> >       set.
> > 
> > BR, YK
> > 
> > >-----Original Message-----
> > >From: ext Ingemar Johansson S
> > >[mailto:ingemar.s.johansson at ericsson.com]
> > >Sent: Monday, July 28, 2008 1:04 PM
> > >To: avt at ietf.org
> > >Cc: Wang Ye-Kui (Nokia-NRC/Tampere); Even, Roni; Randell Jesup
> > >Subject: Re: [AVT] spatial-resolution parameter for RFC 3984bis
> > >
> > >Hi
> > >
> > >
> > >Given the discussion below I really wonder if it is not a better 
> > >alternative to specifiy preferred image sizes using a dedicated 
> > >(generic) SDP attribute.
> > >Even though this is solved in some way in 3984bis we still have to 
> > >provide with a solution for MPEG-4 or H.263 later on.
> > >I drafted a proposed solution for an "a=imageattr" attribute some 
> > >time ago that may be useful
> > >http://tools.ietf.org/id/draft-johansson-mmusic-image-attributes-
> > 01.txt
> > >The draft provides with a solution that may work but lacks a 
> > >requirement spec to outline exactly what the attribute 
> aims for. Also 
> > >extra work is needed to avoid conflicts with related 
> payload format 
> > >parameters.
> > >
> > >Regards
> > >Ingemar
> > >
> > >
> > >=============================
> > >To: <Ye-Kui.Wang at nokia.com>
> > >Subject: Re: [AVT] spatial-resolution parameter for RFC 3984bis
> > >From: Randell Jesup <rjesup at wgate.com>
> > >Date: Mon, 14 Jul 2008 09:35:18 -0400
> > >Cc: roni.even at polycom.co.il, avt at ietf.org
> > >
> > ><Ye-Kui.Wang at nokia.com> writes:
> > >>The resolution must be restricted by the level part of
> > >profile-level-id
> > >>and max-fs. Therefore, if a configuration is agreed (or 
> agreeable), 
> > >>then the answerer must request a resolution under the constraint, 
> > >>and the offerer must accept the request as long as it meets the
> > >restriction.
> > >>This is similar as sprop-parameter-sets, as long as it is 
> complaint 
> > >>with the agreed (or agreeable) configuration, the other side
> > >must accept it.
> > >
> > >I would suggest instead (and did in the thread back in March):
> > >that as a receiver property, it is the *preferred* maximum frame 
> > >size.  The receiver is required to decode and handle any 
> size up to 
> > >the size implied by the level (and max-fs if given), but 
> the receiver 
> > >prefers this size (for example, it may support 1920x1080, but only 
> > >have an LCD display that can display 1200x720, so there's 
> no *need* 
> > >to send it anything higher (and the encoder may prefer to avoid 
> > >encoding more macroblocks, especially in an MCU that shares 
> > >resources)).  In other cases, the sender may have the 
> stream already 
> > >encoded in different resolutions, and this can help select 
> the best 
> > >one to send to save encoding (and bandwidth) resources.
> > >
> > >There is no requirement that the encoder scale, but if it 
> wants to it 
> > >can.
> > >
> > >It also (if I remember) provides aspect-ratio information.
> > >
> > >>Would it work this way?
> > >>
> > >>BR, YK
> > >>
> > >>>-----Original Message-----
> > >>>From: ext Even, Roni [mailto:roni.even at polycom.co.il]
> > >>>Sent: 14 July, 2008 10:38
> > >>>To: Wang Ye-Kui (Nokia-NRC/Tampere); avt at ietf.org
> > >>>Subject: RE: [AVT] spatial-resolution parameter for RFC 3984bis
> > >>>
> > >>>YK,
> > >>>Just to clarify,
> > >>>The text on the receiver side says that this is the requested 
> > >>>resolution, it does not address issue of negotiation, 
> what will the 
> > >>>sender do if it cannot send this resolution, should it 
> reject the 
> > >>>stream?.
> > >>>If this is for negotiation do you except to do a new 
> offer answer 
> > >>>if the receiver wants a new resolution?
> > >>>
> > >>>Roni
> > >>>
> > >>>> -----Original Message-----
> > >>>> From: avt-bounces at ietf.org [mailto:avt-bounces at 
> ietf.org] On 
> > >>>> Behalf Of Ye-Kui.Wang at nokia.com
> > >>>> Sent: Sunday, July 13, 2008 7:57 PM
> > >>>> To: avt at ietf.org
> > >>>> Subject: [AVT] spatial-resolution parameter for RFC 3984bis
> > >>>>
> > >>>>
> > >>>> Concluded from the discussions in the email threads listed
> > >below, I
> > >>>> added a new parameter, spatial-resolution, to the RFC
> > >3984bis draft
> > >>>> (which will be submitted by tomorrow):
> > >>>>
> > >>>> http://www.ietf.org/mail-archive/web/avt/current/msg09425.html
> > >>>> http://www.ietf.org/mail-archive/web/avt/current/msg09650.html
> > >>>>
> > >>>> The parameter is specified as follows (basically as Randell 
> > >>>> suggested in 
> > >>>> 
> http://www.ietf.org/mail-archive/web/avt/current/msg09429.html).
> > >>>>
> > >>>> spatial-resolution:
> > >>>> This parameter MAY be used to indicate the maximum spatial 
> > >>>> resolution of a NAL unit stream or the preferred spatial
> > >resolution
> > >>>> of a receiver.
> > >>>> The value is a base16 [6] (hexadecimal) representation of the
> > width
> > >>>and
> > >>>> height of the spatial resolution, in pixels, separated 
> by a comma.
> > >>>>
> > >>>> Usage in SDP offer/answer:
> > >>>> - When "a=sendrecv" or no direction attribute is used, 
> > >>>> "spatial-resolution" indicates both the maximum spatial 
> > >>>> resolution of the stream sent by the declaring entity and the
> > >preferred spatial
> > >>>> resolution for receiving a stream.
> > >>>> - When "a=sendonly", "spatial-resolution" indicates the maximum
> > >>>spatial
> > >>>> resolution of the outgoing stream.
> > >>>> - When "a=recvonly", "spatial-resolution" indicates 
> the preferred 
> > >>>> spatial resolution for receiving a stream.
> > >>>>
> > >>>> Comments are welcome.
> > >>>>
> > >>>> BR, YK
> > >>>> _______________________________________________
> > >>>> Audio/Video Transport Working Group avt at ietf.org 
> > >>>> https://www.ietf.org/mailman/listinfo/avt
> > >>>
> > >>_______________________________________________
> > >>Audio/Video Transport Working Group
> > >>avt at ietf.org
> > >>https://www.ietf.org/mailman/listinfo/avt
> > >
> > >
> > >--
> > >Randell Jesup, Worldgate (developers of the Ojo 
> videophone), ex-Amiga 
> > >OS team rjesup at wgate.com "The fetters imposed on 
> liberty at home 
> > >have ever been forged out of the weapons provided for 
> defence against 
> > >real, pretended, or imaginary dangers from abroad."
> > >        - James Madison, 4th US president (1751-1836)
> > >
> > >*******************************************
> > >Ingemar Johansson
> > >Senior Research Engineer, IETF "nethead"
> > >EAB/TVP - Multimedia Technologies
> > >Ericsson Research Ericsson AB
> > >Box 920 S-971 28 Luleå, Sweden
> > >Tel: +46 (0)8 4043042
> > >ECN: 850-43042
> > >ECC: 850-43074
> > >Mobile: +46 (0)730 783289
> > >*******************************************
> > >
> 
_______________________________________________
mmusic mailing list
mmusic at ietf.org
https://www.ietf.org/mailman/listinfo/mmusic
_______________________________________________
Audio/Video Transport Working Group
avt at ietf.org
https://www.ietf.org/mailman/listinfo/avt